Académique Documents
Professionnel Documents
Culture Documents
156
0073-1 129/91/0000/0156$01.00 0 1991 IEEE
as independent variables. The model could reasonably As indicated above, using qualitative information to
predict corporate bankruptcy for up to two years in the supplement an investor's forecasting ability in equity
future. Subsequent development and refinement of the markets is beneficial. This type of information is often
model increased the two-year accuracy level and obtained overlooked by investors, perhaps due to its subjective, not
a 70 percent accuracy rate for up to five years in the readily inteqreted. form. While these techniques are
future [2]. A drawback of this linear approach is that it valuable, other methods using non-linear approaches may
classifies some firms as likely to go bankrupt when in further enhance forecasting ability.
actuality, they do not.
Since Alunan's original work, there have been
numemus attempts by researchers to modify and enhance 3. Research Question
his MDA model. Some of the results have been
successful, while others have not. Gentry, Newbold, and
The studies mentioned above have generally
Whitford [lo] found that adding cash-based funds flow
indicated that multiple discriminant analysis, as used in
components to Altman's model provided superior results
the finance discipline, can be a valuable tool to the
in predicting financial failure. They also concluded that
decision maker. It has also been recognized that
cash outflow components were more closely related than
qualitative information can enhance an investor's stock
cash inflow components to corporate failure.
price forecasting ability.
Meyer and fifer's method of predicting corporate
Given the above factors, can non-linear methods
bankruptcy utilized the same financial ratios as were
significantly enhance an MDA model's stock price
found in Altman's model, but added financial data from
predictive power? Specifically, how does multiple
more than one period prior to failure to determine a time
discriminant analysis compare with the neural network
trend. Collins [7] tested both models, using credit union
approach in forecasting stock price performance?
financial data and concluded that this approach adds little,
Additionally. can the use of neural network methods
if anything, to Alunan's model.
enhance the results of a recently published study which
Altman and Spivack compared the Value Line
used multiple discriminant analysis to assess the
relative financial snength system and the zeta bankruptcy
investor's ability to forecast stock price performance.
classification methods of predicting corporate bankruptcy
These questions are addressed in the paragraphs that
[3]. Their findings reveal that although significant
methodological differences do exist, a high correlation follow.
between the methods is evident and that bond systems'
scores correlate well with published bond ratings.
Pinches and Mingo [15] and others utilized the 4. Description of the Data
MDA concept to classify industrial bonds into multiple
categories, using multiple independent variables. Most of As mentioned earlier, qualitative variables can
these attempts have utilized quantitative financial provide an often neglected source of valuable information
variables to construct the model with reasonably good to the investor. Two independent research groups
predictive results. presented the results of the multiple discriminant analysis
Direct applications of the use of the MDA method which used qualitative information found in the
technique to enhance corporate performance are found in fm's annual report to the stockholder 1141 [17]. The
the literature. An example is given by La Flew study conducted by Swales serves as the basis for the
Corporation, who finding financial ruin at their doorstep, application of the neural network approach, which this
worked Altman's model backwards to turn the company paper addresses.
around [4]. The data used in this study was gathered from two
While most of the development of MDA techniques information sources frequently used by investors: The
involves the use of quantitative financial data, other Fortune 500 and Business Week's "Top 1OOO" [18] [191.
approaches using qualitative assessments have been used These soucces provide total return (dividends and stock
in finance. Forecasting how a f m ' s stock will perform price appreciation) and market valuation data,
in equity markets, using qualitative variables found in the respectively, for widely-followed companies. A stock's
fm's annual report to the stockholders, has recently been total return and market valuation are used by investors
attempted. McConnell, Haslem, and Gibson [14] and and financial analysts as performance measures.
Swales [171 have found that qualitative data can provide For this experiment, two separate sets of data were
additional information to forecast stock price gathered. From the Fortune 500 forms, five industries
performance. that offered investors the highest total returns in each year
157
were selected; the samDle consisted of 58 comDanies.
The 10 industries that were reported by Business Week to 5. Neural Network Model
have the highest market valuations provided the data for
the second set; 40 f m s were included in this sample. It
was felt that, if differences across f m s could be found 5.1 Design of the Network
among the top industries, then more pronounced
differences were likely to exist in industries further down The neural network (NN) model is structured in a
the line. four-layered network, as shown in Figure 1: an input
We classified the Fortune set of companies into two layer, two hidden layers, and an output layer. An input
groups according to their total return. Group 1 provided unit has excitatory (positive) or inhibitory (negative)
investors with the highest total returns in their respective connections to a hidden unit in the hidden layer, and a
industries; Group 2 provided the lowest returns in their hidden unit has connections to an output unit in the
industries. We classified the Business Week set of 40 output layer. Therefore, an input unit in this network
companies in each of 10 industries into two groups. structure has indirect connections to an output unit.
Again, Group 1 consisted of those f m s with the highest Input for the network was a list of nine variables:
market valuations for their industries; Group 2 consisted confidence, economic factors outside the fm's control,
of those firms with the lowest market valuations for their growth, strategic gains, new products, anticipated loss,
industries. anticipated gain, long-term optimism and short-term
For each company in the study, the president's optimism. Output was a classification of two patterns: a
letter to stockholders from the annual report for the f m whose stock price performed well and a firm whose
period immediately prior to the group selection year was stock price performed poorly. In a network, each input
studied. A qualitative content analysis technique was parameter is represented in an input unit. Therefore, the
used to classify and tally recurring themes that were network has the nine input units in the input layer and the
identified by similar word or phrases. An examination of two output units in the output layer. The number of
president's letters to stockholders identified nine hidden units necessary to accurately predict the stock
recurring themes commonly addressed in discussion of price performance was determined empirically.
the future. These themes included reference to
confidence, economic factors outside the fm's control,
growth, strategic plans, new products, anticipated losses, Figure 1:
anticipated gains, long-term optimism and short-term Four-Layered Network
optimism.
Each letter was read for content and references to
the themes noted above. The frequency and percentage
of each letter devoted to these themes was then recorded.
When the letter did not contain a specific, direct reference
to one of the themes, a subjective judgment was made by
the researcher as to which, if any, theme should be
credited with the phrase or statement. In those instances
when reference was not made to a theme(s) in the letter,
the data set reflected this finding. The frequency data set
was then used for both MDA techniques and NN methods
to predict stock price performance.
Content analysis techniques have been widely used
in the social sciences 151. Financial researchers have also
Input Layer
applied these techniques to analyze narrative components
of financial reports 161 191 [MI [171. Refer to 1141 1171 x1 x2 x3 ........ x7 x8 x9
for further details of the content analysis technique used
to analyze the president's letter to stockholders.
158
The network for the prediction of stock price performance weights in order to reduce this error. Thus, BPLA is a
uses the following nine input (X - X 9 ) and two output gmhent descent algorithm in which weights in the
,
(Y - Y 2, parameters. network are iteratively modified to minimize the overall
mean square error between desired and actual output
Input Parameters: Output Parameters:
XI: Confidence Y,: well-performing f m s values for all output units over all input patterns. The
amount the weights are to be adjusted for each input
X,: Economic Y,: poorly-perfoming
pattern is determined by the derivative of the error
factors firms function in equation 2 with respect to the weight as
X,: Growth follows.
X,: Strategic plans
X,: New products aE
X,: Anticipated loss
X,: Anticipated gains
A Wji 0~ - aw,, t31
159
four-layered network, performance improved as the Group 2. During the training phase, the MDA model
number of hidden units increased up to a certain point. provided better predictive capability for f m s in the
This supports previous findings on the importance of the lower performance category than for firms in the higher
hidden layer for the different applications [ 111 [201. The performance category. However, during the testing
best performance was achieved by a network with four phase, all models demonstrated better predictive
hidden units on the fnst hidden layer and one hidden unit capability for f m s in the higher performance category.
on the second hidden layer. Increasing the number of
hidden units beyond this point produced no further
improvement, but impaired the network’s performance. Table 1:
Performance of the MDA and Four-Layered Network
on the training and testing data.
Figure 2 Models Training Data Testing Data
Effect of varying the number of Hidden units
in a Four-Layered Network on performance Group Group Mean Group Group Mean
1 2 1 2
Four-
Layer 86% %% 91% 90% 65% 77.5%
I
160
7. Concluding Remarks
Collins, Robert A., "An Empirical Comparison
The study demonstrated that the neural network of Banhptcy Prediction Models,'' Financial
approach is capable of learning a function that maps input
Monagemenf,Summer, 1980, pp. 52-56.
to output and encoding it in the magnitudes of the
weights in the network's connection. The number of
Dum, S. and S. Shekhar. "Bond Rating: A
hidden units employed in the network contributed to its
Non-Conservative Application of Neural
viability. The increase in the number of hidden units
Networks," Proceedings of the IEEE
resulted in higher performance up to a certain point.
International Conference on Neural Networks,
However, additional hidden units beyond the point
1988, Vol 11, pp. 443450.
impaired the model's performance. Comparison of the
NN technique with the MDA approach, indicated that the Frazier, K. b., Ingram, R. W., and Tennyson,
NN approach can significantly improve the predictability
B. M., "A Methodology for the Analysis of
of stock price performance.
Narrative Accounting Disclosures," Journal of
While some limitations of this approach were
noted, it is evident that its use can improve an investor's Accounting Research, Spring, 1984, pp.
decision making capability. Further research into the 3 18-331.
application of neural network techniques. using both
quantitative and qualitative data, is suggested and Gentry, James A., Paul Newbold and David T.
Whitford, " Predicting Bankruptcy: If Cash
encouraged.
Flow's Not the Bottom Line, What Is?,"
Financial Analysts Journal,
REFERENCES September-October, 1985, pp. 47-56.
Altman, Edward I., Robert G. Haldeman and P. Kamijo, K. and Tanigawa. T. "Stock Price
Namyanan, "Zeta Analysis: A New Model to Pattern Recognition: A Recurrent Network
Identify Bankruptcy Risk of Corporations," Approach." Proceedings of the International
Journal of Banking and Finance, June, 1977, Joint Conference on Neural Networks. 1990,
pp. 29-54. Vol 1, pp. 215-222.
Altman, Edward I., and Joseph Spivack. Kimoto, T., Asakawa, K., Yoda, M. and
"Predicting Bankruptcy: The Value LIne Takeoka, M. "Stock Market Prediction System
Relative Financial Strength System vs. the Zeta with Modular neural Networks." Proceeding
Bankruptcy Classification Approach," of the International Joint Conference on
Financial Analysts Journal, Neural Networks, 1990, Vol 1, pp. 1-6.
November-December, 1983, pp. 60-67.
McConnell, Dennis, John A. Haslem and
Ball, Michael, "Z Factor: Rescue by the Virginia R. Gibson, "The President's Letter to
Numbers," Inc., December, 1980. pp. 4548. Stockholders: A New Look." Financial
Analysts Journal, September-October, 1986,
Berlson, B. "Content Analysis in pp. 66-70.
Communication Research," New York: Hafner
Publishing Company, 1971. Pinches. George E., and K. A. Mingo, "A
Multivariate Analysis of Industrial Bond
Bowman, E. H., "Content Analysis of Annual Ratings." Journal of Finance, March, 1977,
Reports for Corporation Strategy and Risk," pp.1-8.
Interfaces,JanuaryFebruary, 19884. pp. 61-71.
161
[161 Rumelhart, D. E., G. E. Hinton, and R. J. [18] "The Fortune 500," Fortune, April 30, 1984;
Williams, "Learning Internal Representations April 29,1985; April 28,1986.
by Error Propagation," In D. E. Rumelhart and
J. L. McClelland [E&), Parallel Distributed [19] "The Top 1OOO: America's Most Valuable
Processing: Exploration in the Microstructure Companies," Business Week,April 18, 1986.
of Cognition. Cambridge: MIT Press, pp.
3 18-362. [20] Yoon, Y., R. W. Brobst, P. R. Bergstresser,
and L. L. Peterson, "A Desktop Neural
[171 Swales, George S., Jr., "Another 'Look at the Network for Dermatology Diagnosis," Journal
President's Letter to Stockholders," Financial of Neural Network Computing,Vol 1, Summer,
Analysts Journal, March-April, 1988, pp. 1989, pp. 43-52.
71-73.
162