82 a E 4.8.5 3:... p z .

 

.a

“PM?

..
r
2.

a

 

.108:
‘2. . v.

.

.. r v
4.9.92.”
7...}...
. .: v

t; UL. .

.. my.
.3: 4 .Ln. 3
.I...I..§. .
ltc'

. .. .

 

 

 

 

 

. 3 :
3.....vexuav
3.: c5... .3
IL ‘4‘.
l

15.; .L .31
new». in. .
rm. 9.3 .0453“,

 

 

 

 

{'8 NHIIHHIHHIIIIllHUI!IIIIHIUUIIIUIIHHIIIIHHUIIHI

19c)? 31293 01810 3790

This is to certify that the

dissertation entitled

USING NON-FINANCIALS AS MEASURES OF
INTANGIBLE ASSETS: A STUDY OF R&D SUCCESSES
IN THE PHARMACEUTICAL INDUSTRY

presented by

Rebecca Toppe Shortridge

has been accepted towards fulfillment
of the requirements for

Ph. D. degreein Accounting

Dr. 19AM}! R. Petroni

 

 

 

Major professor

 

MS U is an Affirmative Action/Equal Opportunity Institulion 0-12771

 

LIBRARY
Michigan State
University

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECAUED with earlier due date if requested.

 

DATE DUE DATE DUE

DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1“ WWW. 14

USING NON-FINANCIALS As MEASURES OF
INTANGIBLE ASSETS: A STUDY OF R&D SUCCESSES
IN THE PHARMACEUTICAL INDUSTRY
By

Rebecca Toppe Shortridge

A DISSERTATION
Submitted to
Michigan State University
In partial fulfillment of the requirements
For the degree of
PHD.
Department of Accounting

1 999

ABSTRACT

USING NON-FINANCIALS AS MEASURES OF
INTANGIBLE ASSETS: A STUDY OF R&D SUCCESSES
IN THE PHARMACEUTICAL INDUSTRY
By
Rebecca Toppe Shortridge

This paper examines the valuation implications of a non-financial measure of
intangible assets in the pharmaceutical industry. Specifically, I hypothesize that the
number of new drugs approved (NDAS) by the Food and Drug Administration provides
incremental information to financial variables when assessing a pharmaceutical
company’s future operating earnings and when predicting its current market values.
Further, the paper hypothesizes that the R&D expenditures of successful product
developers are more valuable than the expenditures of less successfiIl product developers.
The mean number of NDAS obtained by a firm during the sample period is used to
distinguish successful from non-successful product developers.

Tests of the hypotheses are conducted utilizing a sample of 23 traditional
1 pharmaceutical firms and 16 generic pharmaceutical firms. The financial and price data
was collected from archival databases. The FDA data detailing the drug approvals was
obtained from the Food and Drug Administration.

Overall, the results from pooled, cross-sectional regression analyses of the
traditional pharmaceutical firms provide support for the hypotheses. In particular, the
number of approved new drug applications is positively associated with future sales and

with future operating earnings. In addition, the number of NDAS received by a

traditional pharmaceutical firm is correlated with the firm’s stock price three months after
the fiscal year end. Despite the association with stock price, NDAS are not associated
with stock returns. Finally, the results indicate that investors value the R&D of
successful product developers more than that of non-successful developers. These results
suggest that NDAS can be used as a proxy for the knowledge created with investments in
R&D.

The results for the generic pharmaceutical firms are not as strong as those
obtained for the traditional firms. First, NDAS are not associated with sales for the
generic firms although they are marginally associated with operating earnings. However,
NDAS are useful in assessing firm value. These conflicting results might be attributed to
the nature of generic firms. Specifically, generic firms generate revenues by offering
lower prices than branded products. Thus, my tests may might not be sensitive enough to
detect the incremental benefit to annual sales and earnings from one additional drug
product. The results fail to document a relationship between NDAS and stock market
returns. Further, the R&D of successful firms is not valued more positively than the

R&D of non-successful firms in either the price or returns model.

P:
In
he

in

pn

I01

at .'
Slip
10 I.

“h

mac-.-

s
G

I .
l.‘x I

“Has:

ACKNOWLEDGMENTS

I would like to thank the members of my dissertation committee Dr. Kathy
Petroni (chairman), Dr. Joseph Anthony, Dr. Richard Simonds, and Dr. Carl Davidson,
for their time and effort during this process. I would especially like to thank Kathy for
her guidance and prompt attention to my work-—I usually received feedback on a revision
in two days. In addition, working as Kathy’s research assistant during the past four years
has provided me with experiences that made the process much simpler.

I would also like to thank a few other Michigan State faculty members who have
provided me with encouragement and guidance through my tenure at Michigan State. In
particular, Dr. Al Arens and Dr. Matt Anderson have served and will continue to serve as
role models for me during my career as an academic.

My fellow doctoral students have been a source of strength throughout my studies
at Michigan State. Norman Godwin class of 1996, provided invaluable guidance and
support during my first two years as a student. Donna Booker was always there to listen
to my troubles and joys. Lydia Whitt Rosencrants was a constant companion and friend
who provided a shoulder to cry on, a hug for support, and a cheer for my successes.

Finally, I would like to thank my family. My parents, Ron and Judy Toppe,
provided me with the desire to learn, the determination to persevere, and the strength to
know that I could accomplish whatever I wanted to achieve. Brad, my husband, has
made this dream possible. I could not have asked for a more supportive, loving man to
help me through these five years. Finally, thank you, Jackson, for helping me to

understand what matters most in life.

iv

TABLE OF CONTENTS

LIST OF FIGURES ............................................................................ vii
LIST OF TABLES .............................................................................. viii
CHAPTERI.
INTRODUCTION AND OVERVIEW .............................................. 1
1.1 Motivation ............................................................ 2
1.2 Overview of hypotheses ...................................................... 3
1.3 Overview of research design ................................................. 4
1.4 Overview of results ............................................................ 4
1.5 Summary ........................................................................ 7
2. THE PHARMACEUTICAL INDUSTRY .......................................... 8
2.1 Overview of the industry ...................................................... 8
2.2 Drug development and approval ............................................. 9
2.3 Summary ........................................................................ 11
3. THEORY DEVELOPMENT ......................................................... 12
3.1 Economics Literature ......................................................... 12
3.2 Returns to R&D ............................................................... 14
3.3 Non-financials as measures of intangible assets ........................... 16
3.3.1 Using non-financials to assess market performance ............... 16
3.3.2 Using non-financials to predict operating performance ........... 18
3.3.3 Synopsis of papers examining non-financial measures
of intangible assets ..................................................... 19
3.4 Summary ........................................................................ 20
4. DEVELOPMENT OF HYPOTHESES .............................................. 21
4.1 Economic value of knowledge ............................................... 21
4.2 Hypotheses ...................................................................... 23
4.3 Summary ........................................................................ 25
5. SAMPLE SELECTION AND DESCRIPTIVE STATISTICS ................... 26
5.1 Sample selection ............................................................... 26
5.2 Descriptive statistics for full sample ........................................ 27
5.3 Segregated sample ............................................................ 28
5.4 Descriptive statistics for segregated sample ............................... 29
5.5 Summary ........................................................................ 3O
6. CASE STUDY: SCHERING PLOUGH CORPORATION ...................... 32
6.1 Schering Plough Corporation ................................................ 32
6.2 Impact of Claritin on operating performance .............................. 32
6.3 Impact of Claritin on market value .......................................... 34

6.4 Summary ........................................................................ 34

7. EMPIRICAL RESULTS ............................................................... 35
7.1 NDAS: A proxy for knowledge ............................................... 35

7.2 Operating models .............................................................. 37

7.2.1 Tests of hypothesis 1 ................................................... 37

7.2.2 Tests of hypothesis 2 ................................................... 41

7.2.3 Summary of operating models ........................................ 43

7.3 Market models ................................................................. 44

7.3.1 Tests of hypothesis 3 ................................................... 44

7.3.2 Tests of hypothesis 4 ................................................... 49

7.3.3 Summary of market models ........................................... 53

7.4 Sensitivity analysis: Hl-H3 ................................................... 54

7.5 Successful developer models ................................................ 57

7.5.1 Tests of hypothesis 5 ................................................... 58

7.5.2 Tests of hypothesis 6 ................................................... 61

7.6 Summary ........................................................................ 62

8. CONCLUSION ......................................................................... 64
8.1 Review of results ............................................................... 64

8.1.1 Summary of the results for the traditional pharmaceutical firms 64

8.1.2 Summary of the results for the generic pharmaceutical firms 67

8.2 Summary ........................................................................ 67

LIST OF REFERENCES ....................................................................... 69
FIGURES AND TABLES ....................................................................... 73

vi

‘JJ

LIST OF FIGURES

Figure

1 Pharmaceutical R&D timeline ........................................................ 74
2 The knowledge production function ................................................. 75
3 Schering Plough Corporation ......................................................... 76

vii

 

Ta'

LIST OF TABLES

Table

1 Descriptive statistics ..................................................................... 78
2 NDA model .............................................................................. 79
3 Sales model .............................................................................. 80
4 Earnings model .......................................................................... 82
5 Price model .............................................................................. 84
6 Returns model ........................................................................... 87
7 Sales model with patents (Traditional firms only) ................................. 90
8 Earnings model with patents (Traditional firms only) ............................. 92
9 Price model with patents (Traditional firms only) ................................. 94
10 Successful price model ................................................................ 96
11 Successful returns model ............................................................... 99

viii

Chapter One

INTRODUCTION AND OVERVIEW

Recently, the usefulness of Generally Accepted Accounting Principles, GAAP,
has been questioned in the popular press, accounting circles, and academia. Many of the
criticisms focus on the fact that most intangible assets are not measured in financial
statements.1 There is little doubt that a large portion of a company’s value is derived
from its intangible assets. As an indication of this, in 1996 the stock price of firms on the
S&P 500 was four times their asset base (Myers 1996). Numerous research papers have
documented a relationship between intangible asset proxies and stock price (Hirschey
1982, Hirschey and Weygandt 1985, Sougiannis 1994, Amir and Lev 1996, Barth,
Clement, Foster, and Kasznik 1998, Chambers, Jennings and Thompson 1998).

However, U.S. accounting regulations assert that most intangible assets should not be
recognized on a firm’s balance sheet because their values cannot be reliably estimated.2
The International Accounting Standards Committee (IASC) has taken strides to recognize
some intangible assets by issuing a statement that would permit the capitalization of

developed intangibles with identifiable benefits (IASC 1998).

 

IF or a discussion of some of the weaknesses of GAAP see Fox (1996), Davidow (1996), Wallrnan (1996,
1995), Bouwman, Frishkoff, and Frishkoff (1995), AICPA (1994) and Elliott (1992).

2Statement of Financial Accounting Standards 2 requires firms to expense all costs associated with research
and development activities in the period they are incurred (FASB 1974). This rule was established because
the FASB believed it was too difficult to determine the future value of R&D expenditures that might never
result in future sales, let alone a profitable product.

6‘-

$11

rel

Rt

 

 

1.1 Motivation

The purpose of this paper is to examine the valuation implications of a non-
financial measure of intangible assets in the pharmaceutical industry. Specifically, I
consider if new drug applications approved by the Food and Drug Administration provide
incremental information to financial variables when assessing a pharmaceutical
company’s future operating earnings and when predicting its current market value.

The pharmaceutical industry provides an ideal setting to investigate intangible
assets as firms in this industry are dependent on their ability to continually develop new
products for their drug pipelines. To keep the drug pipelines full of new products,
pharmaceutical firms spend millions of dollars on research and development (R&D)
every year. These expenditures are presumably incurred to create an asset that will
increase the firm’s operating performance. However, under US. GAAP, no asset is
recorded for these expenditures.

Numerous accounting papers examine the relationship between R&D
expenditures and market value (Hirschey 1982, Hirschey and Weygandt 1985,
Sougiannis 1994, Aboody and Lev 1998). This research stems from the belief that R&D
creates an intangible asset and thus should be reflected in stock price. However, R&D
expenditures may not be the ideal measure of intangible assets. First, most R&D
expenditures never result in a new marketable product. Further, even if the R&D is
successful, considerable time generally lapses between the occurrence of R&D and the
related product sales. The current paper adds to this literature by examining if the value
of the intangible asset created with R&D might be better assessed with a measure of

R&D success (i.e. new drug applications) rather than with R&D expenditures.

1.2 Overview of hypotheses

In order for a firm to market any pharmaceutical product, it must file a new drug
application with the Food and Drug Administration (FDA). The number of New Drug
Approvals (NDAS) granted to a firm represents the number of products the firm can sell
on the market. R&D expenditures, available to investors in annual reports, only provide
measures of input to the development process. However, much of the investment in
R&D may never result in future products. NDAS, meanwhile, are obtained immediately
before a product will be marketed; and, therefore, NDAs should be a more reliable
measure of future firm profitability than R&D expenditures. Hence, the hypotheses
assert that NDAS are a proxy for the intangible asset created with R&D expenditures and
should be positively associated with future operating performance even after controlling
for R&D expenditures.

Further, if NDAS are associated with future operating performance, they should
also be positively associated with contemporaneous market valuation measures. To test
the market hypotheses, I regress stock price and stock returns on traditional financial
variables, on R&D expenditures, and on the number of NDAS obtained during the fiscal
year.

This paper also examines if the R&D expenditures of successful product
developers are more valuable than the expenditures of less successfirl product developers.
It is likely that some firms are more efficient at generating new drug products than other
firms. A potential measure of a firm’s level of achievement is the number of new drug
approvals received. Thus, using NDAS, I distinguish successful product producers from

non-successful product producers. If the division between firms is appropriate, the R&D

of successful firms should be valued more highly by the market than the R&D of non-

successful firms.

1.3 Overview of research design

Tests of the hypotheses are conducted utilizing a sample of 23 traditional
pharmaceutical firms and 16 generic pharmaceutical firms. The financial and price data
was collected from archival databases. The data detailing the new drug approvals was
obtained from the Food and Drug Administration.3 Regression analysis is used to
examine the hypotheses outlined in the previous section. Several alternate tests are

conducted to ensure the robustness of the results.

1.4 Overview of results

Overall, the empirical results provide support for the hypotheses and imply that
NDAs are a proxy for the intangible asset created with investments in R&D. In
particular, for traditional pharmaceutical firms, the number of approved new drug
applications is positively associated with future sales and future earnings even after
controlling for R&D expenditures. Sensitivity analyses firrther confirm the importance of
R&D success for the traditional firms by showing that NDAS remain positively
associated with future sales and with firture earnings even after including five years of

R&D and patent data.

 

3 Investors and other interested parties may obtain information about product approvals from annual reports,
IO-Ks, or news releases. I chose to obtain the product approvals from the FDA because it is a complete list
of all approvals granted by the Food and Drug Administration.

In addition, the results support the idea that NDAS are important predictors of
market value. For the traditional pharmaceutical firms, NDAS are positively associated
with current stock price, adding $0.34, or just under one percent, to the average share
price of $41.27. Once again, sensitivity analyses confirm the importance of NDAS by
establishing that the coefficient on NDAS is still positive and significant in the price
model afier including five years of R&D and patent data.

Despite the positive association between NDAS and stock price, there is no
apparent relationship between NDAS and stock market returns for traditional
pharmaceutical firms. This is similar to results found in Hirschey, Richardson, and
Scholz (1998) in which patent data is reflected in stock price but is not associated with
firm returns. The lack of significance in the returns model is likely due to the use of a
long window to measure the impact of an event, the receipt of a new drug approval.

The results fiom the tests of the successfirl developer hypotheses for traditional
firms are also insightful. Firms are classified as successful or non-successful developers
based on the mean number of NDAS obtained over the sample period. Firms classified as
successful producers have, on average, higher stock prices than those classified as non-
successful producers. This result supports the conclusion made earlier that NDAs are a
proxy for intangible assets that are not reported on the balance sheets of pharmaceutical
firms. In addition, the R&D of successful firms is positively valued by the stock market
while the R&D of non-successful firms is not. This suggests that investors are
sophisticated as they can distinguish successful from non-successful firms and value firm

R&D investments accordingly. Once again, the returns model does not reveal a

 

 

brz
XI
0p.
the

det

AT
prii
of t
fror
kno
C017

383C

h} firlr
Pred
R&I

P0511

relationship between the proxy for successful drug development and a firm’s stock
market return.

The results for the generic pharmaceutical firms are mixed. First, NDAS are not
associated with sales and are only marginally associated with earnings. This result is
most likely due to the fact that generic firms compete by offering reduced prices on
branded products whose patents have expired. This strategy means that each additional
NDA produced by a generic firm results in a smaller increase in sales revenue and
operating profits than an NDA produced by a traditional firm. Therefore, it is likely that
the incremental benefits obtained from an NDA of a generic firm are too small to be
detected in the annual sales and earnings models.

Even though there is not a strong relationship between operating performance and
NDAS for the generic pharmaceutical firms, NDAS are positively associated with stock
price. More specifically, each NDA increases stock price by approximately four percent
of the average share price of $12.86, or $0.47. Thus, even though the incremental benefit
from an additional NDA cannot be detected in the sales or earnings models, investors
know that NDAs improve the long-range operating performance of the generic
companies and value them accordingly. As with the traditional firms, NDAS are not
associated with market returns.

Finally, generic firms that are classified as successful producers have, on average,
higher stock prices than those classified as non-successful producers. Contrary to
predictions, however, the R&D of successful firms is not valued more positively than the
R&D of non-successful firms. Surprisingly, the R&D of non-successfirl companies is

positively associated with stock price while the R&D of successful companies is not.

 

11163.

Lex

prov

andt

pharr
the (1.
Chap
for St

empir

Once again, there is no relationship between proxy measures of successful R&D and

stock market returns.

1.5 Summary

Taken together, the empirical analyses suggest that NDAS can be used as a
measure of successful R&D efforts. These results are similar to those found in Amir and
Lev (1996), Fomell, Ittner and Larcker (1996), Barth, Clement, Foster and Kasznik
(1998) and Hirschey, Richardson, and Scholz (1998) in which non-financial data
provides supplemental information to financial data in predicting future sales, earnings,
and current stock price.

The remaining chapters are organized as follows. Chapter 2 discusses the
pharmaceutical industry and its drug approval process. A review of relevant literature and
the development of the hypotheses are contained in Chapters 3 and 4 respectively.
Chapter 5 describes the sample used for the study. A case study of a new drug approval
for Schering Plough Corporation is discussed in Chapter 6. Chapter 7 presents the

empirical results. I summarize the paper in Chapter 8.

industry
0116121161

and Du

during
faCt AI
drugs i1
frequer
Resear.
in the 1
{PM}
R&D i
illIIOI‘a'
thfi nee

“Mina

Chapter Two

THE PHARMACEUTICAL INDUSTRY

Before making reasonable assumptions about the important intangible assets in an
industry, it is first necessary to have a thorough understanding of how the industry
operates. This chapter provides an overview of the pharmaceutical industry and the Food

and Drug Administration approval process.

2.1 Overview of the industry

The US. pharmaceutical industry is a thriving business. Statistics indicate that
during the last two decades it has been the world’s leading innovator in drug products. In
fact, American pharmaceutical firms hold patent rights to 92 of the 100 most prescribed
drugs in the US. (Schweitzer 1997). Despite its successful innovations, the industry is
fi'equently criticized for eaming excess returns. However, the Pharmaceutical
Researchers and Manufacturers Association, an organization that follows developments
in the industry, points out that few drugs ever recover their development costs
(PhRMA 1997). The association goes on to argue that without reasonable returns on
R&D investments, the industry would not be able to obtain the capital needed to invest in
innovative medicines that can greatly enhance the lives of many people. To understand
the need for large returns on successful products, the next section discusses the

uncertainties and time involved in the drug development and approval process.

"O

b

CC
EX
DE

ph

the
COI

of.

2.2 Drug development and approval

Prosperity in the pharmaceutical industry is dependent upon a thriving pipeline of
new pharmaceutical products that are created with research and development activities.
As an indication of the significance of R&D activities, in 1995, research-based
pharmaceutical companies spent 19.4% of sales, more than $15 billion, on research and
development (PhRMA 1997). Meanwhile, the average US. industry spent less than 4%
of sales on R&D (PhRMA 1997).

Developing new products to fill the drug pipeline is expensive and involves a
great deal of uncertainty. In fact, statistics show that only one of 5,000 new compounds
ever reaches human testing (Beary 1997). Figure 1, a diagram adapted from Schweitzer
(1997), illustrates the process and length of drug development and testing. The process
begins with a market analysis to determine if a proposed project has economic potential.
This analysis includes an assessment of the pervasiveness of the condition to be treated as
well as the market potential for the new drug. After a project passes this initial market
analysis, the next phase involves research and testing to determine what causes the
condition. This research, frequently called basic research, is the longest and most
expensive part of drug development. The focus of this phase of research is to discover a
new chemical that will enhance the treatment of the condition. A new chemical in the
pharmaceutical industry is referred to as a new molecular entity (NME).

Once basic research has discovered a new molecular entity, testing begins with
the assistance of computer models and cell cultures. If this step is successfirl, testing
continues on animals. These initial development and testing procedures take an average

of 42.6 months to complete (Schweitzer 1997). Many firms, upon discovering an NME,

 

 

apply 1

meN\

rqmm
drug w
mare
Rdmy
meah
platebt
These!
dhhe
process

Meme

06pm

3 mm

apply for patent protection. A patent grants the filing firm the exclusive right to market
the NME for seventeen years from the date the patent is approved.4

After completing pre-clinical testing, the NME must pass the rigorous testing
requirements of the FDA. First, a firm files an application for an investigational new
drug which allows the company to begin testing on human subjects. Human testing
proceeds in a series of three phases. Phase I trials typically include a small number of
healthy volunteers. Phase II trials are performed on individuals who are suffering fiom
the ailment the drug was developed to cure. These tests are conducted using double-blind
placebo testing. Finally, Phase III involves testing in hospitals and outpatient settings.
These tests establish the drug’s safety and effectiveness. If the proposed product passes
all three phases, the firm then files a new drug application with the FDA.5 This entire
process, from the time early research begins until the time a new drug is approved,
averages approximately 12 years (Schweitzer 1997).6

Although measures of innovative input, R&D expenditures, are available early in
the product development cycle, they may have little meaning in determining the value of

a company because of the time and risk of getting a product to market. Therefore, output

 

4In 1995, the international World Trade Organization enacted a treaty extending the patent period for all
member countries to 20 years from the date of filing. However, it is unclear if this treaty actually extends
the period of exclusive marketing as the clock starts at the date of filing instead of the date of patent
approval. (Schweitzer 1997)

5 It is not unusual for a firm to seek product approval in a foreign country before seeking approval in the
United States. This occurs because U.S. regulations are much more stringent that those in many foreign
countries. Hence, a drug may be marketed for several years in a foreign country before it is available in the
United States.

“The process described is frequently abbreviated for drug products that copy existing chemicals. For
example, generic pharmaceutical firms copy off-patented drugs. Because the generic firms are duplicating
already marketed drugs, these firms merely must establish that the copied product is chemically identical to
a drug that has already been approved.

10

descrip
the cha
llIe rer

the hp

measures, such as successful product developments, may be more useful to investors

when assessing the value of a firm.

2.3 Summary

This chapter provides a brief overview of the pharmaceutical industry as well as a
description of the drug approval process. One of the most important ideas established in
the chapter is the risk and time involved in the drug development and approval process.
The remaining chapters of the paper rely on the industry knowledge to develop and test

the hypotheses outlined in Chapter 4.

ll

 

etonor
produt‘
expent
proxie

indepe

3.1

311d m:
papers
manuf;
Patents
bOIh SC
011me
sud}; ;

industr

a more

“(Mia

Chapter Three

THEORY DEVELOPMENT

There are three different research streams that are relevant to the current project:
economics literature that examines the relationship between R&D efforts and R&D
productivity, accounting literature that examines the relationship between R&D
expenditures and stock prices, and accounting literature that examines non-financial
proxies for intangible assets. These three streams of literature will be discussed

independently in Sections 3.1, 3.2, and 3.3.

3. 1 Economics Literature

Relevant economic papers have focused on the relationship between R&D efforts
and measures of R&D productivity. Bound et al (1984) is one of the first economics
papers to evaluate the relationship between R&D and patents. Using a sample of all
manufacturing firms listed on the 1976 Compustat Tapes that report both R&D and
patents, Bound et a1 plot the log of patents against the log of R&D expenditures,
both scaled by assets. The plot indicates that there is a strong correlation between the
number of patent applications and R&D expenditures. Particularly relevant to the current
study, a regression of patents on R&D that controls for industries indicates that the drug
industry has a significantly higher than average propensity to patent.

Pakes and Griliches (1984) confirms the relationship between patents and R&D in
a more comprehensive study that uses patents as a measure of economically significant

knowledge. Using the data from Bound et al (1984), Pakes and Griliches show that when

12

a firm changes its R&D expenditures, parallel changes occur in its level of patenting. In
addition, they attempt to estimate the lag structure of R&D. Although the lag effects are
significant, the contemporaneous relationship between R&D and patents appears to be
most dominant.

Hall, Griliches, and Hausman (1986) expands the sample used in Pakes and
Griliches (1984) and focuses on establishing the lag structure between patents and R&D.
Their model, like that of Pakes and Griliches’, assumes that patents are an indicator of
R&D successes. Their results again confirm that there is a strong contemporaneous
relationship between R&D expenditures and patents even after controlling for firm size
and the lag structure of R&D. They also show that the R&D history included in their
model (they include as many as eight years of lagged data) seems to add little to the
explanation of the current year’s patent applications. However, Hall, Griliches, and
Hausman point out that it is difficult to correctly estimate the lag structures because the
R&D expenditures across time are highly multicollinear.

These three economic papers all reach the same basic conclusion—there is a
strong contemporaneous relationship between patenting and current R&D expenditures.
While the authors of all three of these papers acknowledge that they believe R&D history
should be important to the prediction of patents, they cannot establish a strong
relationship between the R&D lag structure and patents because of multicollinearity. The
results from these three papers do suggest that patents are an indicator of the success of

R&D activities.

13

 

testin
asset.
value
mule

anal)

Variz
the r

15a:

\‘alu

firm

be.

3.2 Returns to R&D

Some accounting papers attempt to measure the benefits of R&D investments by
testing if the expenditures create an intangible asset. To establish the presence of the
asset, R&D expenditures are shown to be associated with operating profits and firm
value. For example, Hirschey (1982) examined the impact of intangible assets on the
market value of a sample of 390 firms listed on the 1977 Fortune 500. Hirschey first
analyzed the issue from a theoretical perspective by creating a model that showed that,
with a constant rate of amortization and expenditure growth, the relationship between the
capital created by intangibles and current levels of expenditures is proportional. Relying
on this proportional relationship, Hirschey developed a model to establish the market
value of a firm by relying on indicators of firture profit, including: current profit, book
value of tangible assets, R&D expenditures, and advertising expenditures. Using the
sample fi'om Hirschey (1982), Hirschey and Weygandt (1985) altered the above model
slightly by using Tobins’ Q as the dependent variable. This measure of the dependent
variable accounts for the replacement costs of tangible assets, thus more directly isolating
the market value of intangibles. The empirical results of both papers indicate that R&D
is a significant predictor of the current market value of the firm.

More recently, Sougiannis (1994) and Lev and Sougiannis (1996) examine the
value of corporate R&D and its capitalization. Sougiannis (1994), using a sample of 573
firms listed on the National Bureau of Economic Research’s RND-Panel, first shows that
R&D is positively associated with earnings by regressing earnings on R&D expenditures
lagged up to seven years with various control variables. The R&D coefficients vary from

positively to negatively significant. This suggests that in some years, the cost of R&D

l4

 

oumfl
that a c

dollars

estimat
Sougiar
one dol.
both pa;
hnancia

R&D.

Sougianr
of R&D

because r
involved

the intang
PIOblem (
"311313165 ;

discussed

outweigh the benefits. He then includes R&D in a valuation model. The results show
that a one-dollar increase in R&D results in a total increase in market value of five
dollars.

Lev and Sougiannis (1996) further extends prior work by including industry-wide
estimates of the value of R&D. Particularly relevant to the current project, Lev and
Sougiannis’ results suggest that for firms in the chemical and pharmaceutical industry, a
one dollar investment in R&D increases future earnings by $2.63. Overall, the results of
both papers imply that R&D is important to investors and that investors appear to restate
financial statements when determining stock price to include a measure of capitalized
R&D.

All four of these papers, Hirschey (1982), Hirschey and Weygandt (1985),
Sougiannis (1994) and Lev and Sougiannis (1996), rely on financial statement measures
of R&D expenditures to assess the value of intangible assets to the market. However,
because of the length of time required to produce a new product and the uncertainty
involved in the development process, R&D expenditures may not be the best measure of
the intangible asset. Some recent accounting researchers have begun to address the
problem of using expenditures as a measure of intangible assets by using non-financial
variables as measures of the intangible assets. Several of these accounting papers are

discussed in the next section.

15

 

atten
samp
111 III:
this. .
be us.
finant
cellul.
geogr.
explai
comp;
Share ;
SUgges

to east

“7'3 tStIr:
mim :1

3.3 Non-financials as measures of intangible assets
3.3.1 Using non-financials to assess market performance

Amir and Lev (1996) is the one of the first papers in a new line of research that
attempts to use non-financial information as an indicator of intangible assets. Their
sample is comprised of firms in the wireless communications industry. Most of the firms
in their sample are start-up companies that have never had positive earnings. Because of
this, Amir and Lev believe that their sample firms’ financial information is not likely to
be useful in predicting stock price or stock returns. Indeed, their research shows that
financial information alone is of little use in predicting stock prices and returns in the
cellular communications industry. Meanwhile, POPS, a measure of the population in
geographic areas were the firm is licensed, is a significant predictor of price. The authors
explain this result because POPS is a measure of the future grth potential of the
company. In addition, when POPS is included in the valuation model, book value per
share and earnings per share are positively associated with stock price. These results
suggest that financial and non-financial information provide complimentary information
to each other.

Barth, Clement, Foster, and Kasznik (1998) examines the importance of brand
values to the capital market. Barth et al rely on estimates of the dollar value of various
brands from an annual survey published in Financial World.7 These estimates make their

study of non-financial measures unique because the authors have an estimate of the dollar

 

7The estimates include an assessment of brand-related profits (operating income from a brand minus
operating income from a similar generic product), leadership in the market, stability in the market, and
others. This process produces a dollar amount estimate of the value of the brand. For example, the Gillette
name was estimated to be worth $10.3 billion.

16

 

valu
regr
ralu

Ialu

kai
reha
toti

and

non-
usin
indi
not .
the;
fitnr
pate
preC
{IOn

Chat

’Ths

\

a hit

:(Ifm.
T‘E'i .
¥H 1“;

value of the intangible instead of a count variable as used in most papers. The authors
regress market value on book value per share, earnings per share, and the total brand
value per share. Using a fixed-effects pooled regression, they demonstrate that brand
values are positively associated with stock price, even after controlling for net income
and the book value of equity. Thus, the results are counter to arguments asserted by
SFAS 2 which does not allow recognition of intangible assets because they cannot be
reliably estimated. This study may be biased, however, because the brand value assigned
to the firms might be partially determined by market prices. If this is true, brand value
and stock price are associated by construction.

Hirschey, Richardson and Scholz (1998) considers the valuation implications of
non-financial measures as proxies for the intangible asset created with R&D expenditures
using a sample of high-technology firms.8 In particular, the authors posit that patents are
indicators of the intangible asset created by inventive activities. However, patents are
not a perfect measure as they vary in economic value and scope. Patent scope determines
the ability of competitors to manufacture sister products that can take away market share
from the original developer. Thus, the authors argue that measures of the quality of
patents obtained may provide additional incremental information to the market when
predicting stock price. To test the theory, they obtain three measures of patent quality
fiom the T E CH-LINE database: frequency of patent citations, number of scholarly

citations on the patent application, and the median age in years of earlier US. patents

 

8The sample is comprised of firms listed in the TE CH-LINE database. Non-manufacturing firms and those
which are in low-tech industries are eliminated from the sample. Thus, the final sample is comprised of
firms in the following industries: chemicals and allied products; petroleum refining; industrial and
commercial machinery and computer equipment; electronic and other electrical equipment; transportation
equipment; measuring, analyzing and controlling instruments; and communications.

17

referer
Richar
price a
grantet
expenc
indit‘id
increm
include

patent c

3.3.2

intestig
Americ.
satisfac‘
relation
Pearson
Perform
moms:
Ififlecte‘

aCCOUm

referenced in the patent application. Relying on the Ohlson model (1995), Hirschey,
Richardson, and Scholz run a pooled time-series, cross-sectional regression using stock
price as the dependent variable. The independent variables are the number of patents
granted to firms, the three measures of the quality of patents, book value, earnings, R&D
expenditures, and advertising expenditures. Their regression results show that,
individually, the number of patents as well as the three measures of patent quality provide
incremental information to financial data when assessing stock price. When jointly
included in the valuation model, the number of patents and two of three measures of
patent quality are significantly positive.
3.3.2 Using non-financials to predict operating performance

In a somewhat different fiamework, Fomell, Ittner, and Larcker (1996)
investigate the valuation consequences of customer satisfaction indexes. Using the
American Customer Satisfaction Index (ACSI), they first evaluate if customer
satisfaction is reflected in contemporaneous financial statements. To examine the
relationship between customer satisfaction and financial statements, the authors provide
Pearson correlations between ACSI scores and financial ratios that reflect operating
performance. The correlations show that the ACSI scores have a statistically positive
association with return on assets, implying that the ACSI scores are at least partially
reflected in accounting returns. After concluding that ACSI scores are reflected in
accounting numbers, the authors continue the analyses to assess if customer satisfaction
is economically relevant to the stock market. Correlations between the ACSI scores and
stock valuations (price earnings and market-to-book) are also significantly positive. To

strengthen the results, the authors perform regression analyses of the two stock valuation

18

 

mEJSl
COIIIII
C8565.

signit‘

(1997’
partiCi
intang
airline
on-tirr
associ
pOSlllt
63min
author
marke
FOrne‘;
in part
the air

33.3

IncaSu:

and La

measures on the ACSI scores, risk (beta), Value Line’s estimates of dividends per
common share, and either forecasted earnings per share or forecasted book value. In both
cases, the estimated coefficient on the customer satisfaction scores is positive and
significant.

In a paper similar to that of Fomell, Ittner, and Larcker (1996), Behn and Riley
(1997) examine the usefulness of non-financial indicators to decision makers. In
particular, the authors examine the association between non-financial indicators of
intangible assets and financial performance measures. Using data from the US. domestic
airline industry, Behn and Riley use an instrumental variables approach to establish that
on-time performance, mishandled baggage, and in—flight service are significantly
associated with customer satisfaction. They then use regression analysis to document a
positive relationship between the intangible asset, customer satisfaction, and operating
earnings and revenues. To reduce the possibility of correlated omitted variables, the
authors are careful to include additional key industry variables in the model: load factor,
market share, and available miles scheduled. Their results, similar to those found in
Fomell, Ittner, and Larcker (1996), establish that non-financial performance information,
in particular customer satisfaction, can be useful in predicting financial performance in
the airline industry.
3.3.3 Synopsis of papers examining non-financial measures of intangible assets

The results from Amir and Lev (1996), Barth, Clement, Foster, and Kasnzik
(1998), and Hirschey, Richardson, and Scholz (1998) indicate that non-financial
measures of intangible assets are positively associated with market value. Fomell, Ittner,

and Larcker (1996), and Behn and Riley (1997) provide evidence that suggests that non-

19

 

kno

the

WP

pert

aIE

R8:

financial information can be used to predict future operating performance. Overall, both
groups of papers imply that financial and non-financial measures are complimentary to
each other and that both should be utilized by decision makers when assessing a firm’s

market value and when predicting a firm’s future operating performance.

3.4 Summary

There is little doubt that the research and development activities of firms create
knowledge that can be used to create a competitive advantage. What is questionable is
the best way to measure the value of that knowledge. In this regard, numerous research
papers have examined the relationship between R&D expenditures and various
performance indicators. The disadvantage of this methodology is that much R&D may
be wasted and that fruitful R&D may be incurred years before the benefits of the R&D
are realized. Thus, using non-financial measures of the knowledge obtained through
R&D efforts in conjunction with the financial measures of R&D expense may enhance

the prediction of operating and market performance measures.

20

int.

dit

4.1

val

me

she

€211

Chapter 4

DEVELOPMENT OF HYPOTHESES

Pakes and Griliches (1984) lends insight into the problem of valuing the
intangible asset created by R&D expenditures. Their insight is used to unite the three

diverse streams of literature presented in Chapter 3 and to develop the hypotheses.

4.1 Economic value of knowledge

Pakes and Griliches (1984), argues that R&D is used to create economically
valuable intellectual knowledge, K (See Figure 2). A problem arises because K, a
measure of the productivity of R&D activities, is not observable. Thus, in Figure 2, K is
shown as k, the net accretion of valuable knowledge. As discussed in Chapter 3,
economists working in this area have generally tried to measure k with the number of
patents applied for or granted to firms. However, there are numerous problems with
using patent counts as a measure of economically significant knowledge in the
pharmaceutical industry. First, pharmaceutical firms typically apply for patents very
early in the drug development process. DiMasi et al (1991) and Dranove and Meltzer
(1994) suggest that a twelve to fourteen year delay between patenting and FDA approval
is not unusual. Further, in order for a pharmaceutical firm to market a new drug, it must
pass the rigorous testing requirements of the Food and Drug Administration. A patent
provides no information about the ability of a new chemical to pass FDA testing; it
merely indicates the existence of a new compound and grants production rights to the

developer. Toole (1997) argues that patents provide no indication as to the therapeutic

21

value I
pharm:
retreat
(199.7).
that uh

to tirm

R&D su
Specific
measure
Very nea
count of
uncertair
eliminate-
products

measure

value of a chemical. In addition, some firms choose not to patent. Generic
pharmaceutical companies have few, if any, patents because their business tactic is to
recreate and produce off-patented products at low prices. And, as discussed in Toole
(1997), many innovations are not patented because firms rely on the old adage: “patent
that which you cannot keep secret.” Finally, it is difficult to relate the number of patents
to firm performance as the economic significance of individual patents varies immensely.
To alleviate some of the problems associated with using patents as a measure of
R&D success, I rely on a proxy that is unique to the pharmaceutical industry.
Specifically, the number of new drug applications approved by the FDA will be used as a
measure of successful R&D activities.9 NDAS are obtained by a pharmaceutical firm
very near the time that a new drug product is introduced on the market. Therefore, a
count of NDAS should be more reliable than patent counts because much of the
uncertainty with product viability and all of the uncertainty of FDA approval has been
eliminated. Further, all firms, including generics, must obtain NDAS before their
products can be introduced on the market. Thus, the number of NDAS can be used to

measure the productivity of R&D, regardless of firm type.

 

9Another potential measure of a successful product is New Chemical Entities (NCES). This measure is
used in Toole (1997) to assess the impact of federally funded basic research on innovation in the
pharmaceutical industry. NCEs are granted to completely new molecular entities and not to those that
replicate existing entities. Thus, drugs that slightly improve or that copy an existing chemical would not be
included as a new product. Further, generic drug companies would appear to have no marketable products
as they do not create new chemical entities. Thus, using NCEs as a measure of successful products would
greatly limit the measure of a successful product. Therefore, I do not rely on NCEs as a measure of new
product developments. The relationship between NCEs and NDAS might be interesting to examine in
future research.

22

art
CC

sh

be

ole:
Surf.
if or
0D 3

4.2 Hypotheses

The top part of Figure 2 shows that the knowledge created by R&D activities is an
important predictor of a firm’s future economic success; it helps to create future earnings
and increases the market value of the firm. These economic successes, the 25 in Figure 2,
are ultimately the variables that are interesting to predict. If NDAS are measures of
economically valuable knowledge created by successful R&D projects, then NDAS
should contribute to the future operating performance of pharmaceutical firms.lo
Hypotheses 1 and 2 state the predicted relationships between NDAS and future operating

benefits:

H1: The number of new drug applications approved by the Food and Drug
Administration during fiscal year t-I will be positively associated with

the sales revenue of pharmaceutical firms during year t.

H2: The number of new drug applications approved by the Food and Drug
Administration during fiscal year t-I will be positively associated with

the operating earnings of pharmaceutical firms during year t.

If NDAS are available to decision makers and if NDAS are associated with future

operating earnings, they should also be associated with current stock prices and market

 

'0 Perhaps the ideal measure of the knowledge created with R&D activities would be a measure of the value
of each new drug application. It is likely that some new products generate much higher sales than others
and thus result in higher earnings. For example, Claritin is a widely prescribed product as many people
sufi'er from allergies. However, Pitocin, a drug that is used to induce labor in pregnant women, would not
be broadly prescribed. This type of data is difficult, if not impossible, to obtain. Therefore, I simply rely
on a count of NDAS as a rough proxy for the value of knowledge obtained through R&D expenditures.

23

ICIUI'Ilf

variah

benefit
0f thin,
film pr

Other ir

IElatirc
Precise
R&D tl
idea. 1,

huge” n

returns. Hypotheses 3 and 4 state the expected relationship between NDAS and market

variables:

H3: The number of new drug applications approved by the Food and Drug
Administration during fiscal year t will be positively related to the firm ’5

contemporaneous stock price.

H4: The number of new drug applications approved by the Food and Drug
Administration during fiscal year t will be positively related to the firm ’s

contemporaneous market return.

Finally, the X’s in Figure 2 represent other observed variables that influence firm
benefits. They can be thought of as control variables in the relationship and might consist
of things such as tangible assets and firm size. The v’s are other variables which impact
firm profitability but that cannot be directly observed. These might be comprised of
other intangible assets such as customer satisfaction or name recognition.

It is possible that the number of NDAS can be used to identify firms which are
relatively more successful at creating new products for their drug pipeline. More
precisely, it is likely that some pharmaceutical firms obtain more NDAS per dollar of
R&D than other pharmaceutical firms. Conversations with market analysts confirm this
idea. In fact, the analysts suggest that some pharmaceutical firms are more methodical at

targeting their research efforts and thus are more successful at creating new products. If

24

(1’)

8|

0.

(1.

ill;

Sit)

this is true, the R&D of successful firms should be more highly valued by the stock

market than the R&D of non-successful firms.ll Hypotheses 5 and 6 explore this idea:

H5: The research and development expenditures of successful firms will
have a greater positive association with stock price than the research

and development expenditures of non-successful firms.

H6: The research and development expenditures of successful firms will
have a greater positive association with market returns than the

research and development expenditures of non-successful firms.

4.3 Summary

Pharmaceutical firms spend millions of dollars on research and development in an
effort to create knowledge that is economically valuable to the firm. It is not possible to
directly observe knowledge or the value it creates. This chapter asserts that the number
of new drug approvals received by pharmaceutical firms can be used as a proxy for
economically valuable knowledge. Altematively, the number of new drug approvals can
be used to distinguish successful product developers from non-successful product
developers. The R&D of successful developers should be more valuable than that of the

non-successful developers. These hypotheses will be tested in the forthcoming chapters.

 

1'Based on results obtained in prior research (Hirschey (1982), Hirschey and Weygandt (1985), Sougiannis
(1994) and Lev and Sougiannis ( 1996)), I assume that R&D expenditures are positively associated with
stock price for all firms. If this is true, the R&D of successful firms should have a larger positive
association with stock price than the R&D of non-successful firms.

25

measur
onR&.
industr
approx:
produc

selectic

5.1

code of
foreign
Pfimar)
Finally.

PTOCCdL

Chapter Five

SAMPLE SELECTION AND DESCRIPTIVE STATISTICS

The pharmaceutical industry provides a good arena to examine non-financial
measures of intangible assets as this industry spends a relatively large amount of money
on R&D compared to that spent by other industries. Further, the regulation of the
industry makes it possible to find a publicly available proxy, the number of new drug
approvals obtained from the Food and Drug Administration, to represent successfiil
product development. This chapter includes a detailed description of the sample

selection process and descriptive statistics for the final sample of firms.

5.1 Sample selection

The initial sample is comprised of 186 firms in the pharmaceutical industry, SIC
code of 2834, included on the 1997 PC version of Compustat. From this sample, 19
foreign firms were eliminated.12 Additionally, 31 firms were eliminated because their
primary business activity is not consistent with the pharmaceutical industry.13
Finally, 17 firms were excluded because of data availability.” These elimination

procedures resulted in a sample of 119 firms.

 

l2Two of the 19 frrrrrs are based in the US. but only sell products in foreign countries.

'3 For example, Jones Medical Industries was eliminated because it is exclusively a distributor of
pharmaceutical products. Polymedica Industries was eliminated because it is primarily involved in
manufacturing medical equipment.

l"The large majority of these omissions (14) occurred because the frrm was not included in the CRSP
database. Three of the exclusions were for firms that had only one year of data on Compustat. For several
models, lagged variables are used. Thus, each finn was required to have at least two years of data to be
included in the sample.

26

I
the twelt
C RSP. I
From 193
the NDA
with the;
Standard
to foreigr
Records
procedur
1,667) 0:

sample.

5.2 I

all Oldie
firms arr
brim

$1.4 bill
amanger
billion_

“if Sar

fame or

“at"? data

I collected all financial statement data for the sample firms from Compustat for
the twelve year period from 1985 through 1996. Price and return data was obtained from
CRSP. The number of NDAS was obtained from the Food and Drug Administration.
From 1985 to 1996, 4,997 New Drug Applications were approved by the FDA. Many of
the NDAS are granted to subsidiaries of the sample firms. In order to match the NDA
with the parent firm, I used the cross-listing of parent and subsidiaries contained in
Standard and Poor’s Corporation Records. In addition, several of the NDAs are granted
to foreign or private firms; these were identified using Standard and Poor’s Corporation
Records and Wards’ Business Directory of US. Private and Public Companies. These
procedures allowed me to match 3,739 of the 4,997 NDAS. Approximately 45% (or
1,667) of the identified NDAS relate to the 119 pharmaceutical firms included in the

sample.

5.2 Descriptive statistics for full sample

Table 1, Panel A, provides descriptive statistics for the 695 firm-years for which
all of the relevant data was available.15 The descriptive statistics show that the sample
firms are very diverse. Assets range from a low of $0.4 million to a high of over $24
billion. The sample is skewed as the median assets are $47 million while the mean is
$1.4 billion. Total sales, which consists of sales as well as fees from licensing
arrangements and royalties, ranges from $0 to $21.6 billion with an average of $1 .2

billion. The firms with no sales are generally small, start-up companies that are spending

 

lsThe sample is comprised of 119 firms during the 12 year period from 1985 to 1996 for a possible testable
sample of 1,428 firm-years. However, numerous firms, especially the small, development firms, did not
have data for all of the years. Therefore, the final sample is comprised of 695 firm-years.

27

large amounts of money to develop their pharmaceutical pipelines but have yet to
produce a successful product. R&D expenditures range from $0 to $1.9 billion with
a mean of $124 million. Thus, on average, the firms in this sample are spending 10
percent of their total revenues on research and development. Finally, these firms, on
average, obtain between two and three NDAS per year. Again, this figure is highly
variable with most firms obtaining no NDAS in a given year and one firm obtaining a

high of 64 NDAS in one year.

5.3 Segregated sample

After examining the business descriptions and financial data of the firms, it was
apparent that the sample is actually comprised of three distinct groups: traditional firms,
development firms, and generic firms. Traditional firms tend to be large, well-
established firms with numerous well-known products. All firms that are members of
PhRMA were classified as traditional companies.16 Several additional firms whose
business descriptions indicated that they manufacture original drug products were added
to this group. None of the traditional firms incur R&D expenditures greater than their

revenues. Twenty-three of the sample firms were classified as traditional firms.'7

 

l"PhRMA states that member firms are comprised of companies “significantly engaged in the manufacture
and marketing of finished dosage form ethical pharmaceutical or biological products under [its] own brand
names and significantly engaged in pharmaceutical, biopharrnaceutical or biological research and
development of new molecular entities or new therapies and who will continue to conduct such research
and development.” (PhRMA 1997) Merck and Eli Lilly are examples of firms assigned to this group.
l7Twenty of the twenty-three traditional firms obtained at least one NDA during the sample period. Two
firm’s annual reports indicate that it has at least one drug on the market with FDA approval. However, 1
was unable to identify the NDA. This could occur if the NDA was obtained prior to 1985 or if the
company has a subsidiary or corporate partner that was not identified. The final company markets several
products that were purchased from other firms. Further, it has applied for additional NDAs but none were
granted between 1985 and 1996.

28

In contrast to the traditional pharmaceutical firms, development firms tend to be
small firms with little or no revenues but very large R&D expenditures. Development
firms usually focus on creating cutting-edge products. These firms frequently identify
themselves as biotechnology companies. Many of these companies have less than 20
employees, most of whom are scientists. Few of these firms have the capability to
manufacture the products they develop and usually license any developed products to
traditional firms. This strategy allows them to focus on their expertise—research and
development of new drug products. Most of the firms assigned to this group are listed in
the North American BioTechnology Directory.18 The development group is comprised of
eighty firms, nearly two-thirds of the sample.

Finally, generic pharmaceutical firms differ from both traditional and
development firms because they do not conduct original research. Instead, generic
companies copy successful off-patent products originally created by traditional or
development firms. The 16 firms included in this class specifically indicated in their
company annual reports that their primary operating focus is the production of generic

pharmaceutical products.

5.4 Descriptive statistics for segregated sample
Descriptive statistics for these three different categories of firms are included in
Table 1, Panels B, C and D. As expected, traditional firms, on average, are much larger

than either the development or generic firms with mean assets of $4.3 billion compared to

 

I8Several traditional pharmaceutical firms are also listed in the North American BioTechnology Directory
(1995) as they conduct biotechnology research. However, they have much more in common with
traditional firms than with development firms and thus were only included as traditional firms.

29

534 mi
usually
Mean“
percent.
spend. <
incur re
Finally.
consiste
resource
uorking
hate liu

less on I

R&Das

the FDA
aC‘lUired

r.)-
«In
(I)

IChOSE 1(

dEVelopn

Chaplfl h

$34 million and $127 million respectively. Both traditional firms and generic firms are
usually profitable with average earnings of $559 million and $7 million respectively.
Meanwhile, most development firms incur losses. It is insightful to compare R&D as a
percentage of revenues across the three classes of firms. Specifically, traditional firms
spend, on average, 10 percent of their revenues on R&D. Meanwhile, development firms
incur research and development expenditures that are twice as large as their revenues.
Finally, generic firms only spend 7 percent of their revenues on R&D. This statistic is
consistent with expectations. Traditional pharmaceutical firms commit a large amount of
resources to R&D but are usually profitable. The development firms, however, are
working to establish new products. Thus, they spend most of their capital on R&D but
have little or no revenues to cover the cost of their investments. The generic firms spend
less on R&D as they are focused on copying existing successful products. Hence, their
R&D as a percentage of sales is lower than both the traditional and development firms.

All of the hypotheses focus on the number of new drug applications approved by
the FDA. However, less than 1% of the observations for the development firms have
acquired one or more NDAS. Therefore, I do not test the hypotheses on the 80

development firms.

5.5 Summary

The sample selection procedures resulted in a group of 119 pharmaceutical firms.
I chose to separate these firms into three distinct classifications: traditional firms (23),
development firms (80), and generic firms (l6). Descriptive statistics presented in the

chapter highlight the differences between the three classes of firms. Because the

30

 

det‘elopf

Chapter

development firms have very few new drug approvals, the hypotheses are tested in

Chapter 7 using only the traditional and generic pharmaceutical groups.

31

useful tr
periorrr
pharrna

to sales

6.1

of 55.?
product
cardiox-
allergy
marked
Figure

Stilton;

Chapter Six

CASE STUDY: SCHERING PLOUGH CORPORATION

Before presenting the results from an industry-wide test of the hypotheses, it is
useful to examine the impact of an NDA on one company’s operating and market
performance. This in-depth review of one company provides institutional detail for the
pharmaceutical industry, presents an example of the time-line from a new drug approval

to sales, and fumishes anecdotal support for the hypotheses to be tested.

6.1 Schering Plough Corporation

Schering-Plough Corporation is a traditional pharmaceutical firm with 1996 sales
of $5.7 billion and total assets of $5.4 billion. The company concentrates on developing
products in the allergy/respiratory, anti-infective/anti-cancer, dermatological, and
cardiovascular lines. Claritin, a non-sedating antihistamine, is a premier product in the
allergy/respiratory line as it is the world’s largest-selling antihistamine. Claritin received
marketing clearance in the US. from the Food and Drug Administration in April of 1993.
Figure 3 provides a graphic presentation of the impact of Claritin’s U.S. approval on

Schering Plough Corporation in terms of sales, net income, and stock price.

6.2 Impact of Claritin on operating performance
Before the 1993 US. approval, Claritin was marketed in foreign countries.
However, the US. approval had a tremendous impact on the operating performance of

Schering Plough. In 1991 and 1992, world-wide sales of Claritin (all outside of the US.)

32

hover
5130
mark
sales
com;

1993

full 3
Son
mill
entir

the :

peri
the r
anal
com
With
Sale:

Prod

hovered around $100 million. After FDA approval in April of 1993, Claritin generated
$130 million in US. sales over an eight month period. In 1994, the first full year of US.
marketing, Claritin achieved world-wide sales of $505 million and pushed total company
sales to $4.7 billion. Further, as can be seen from Figure 3A, the increase in Claritin sales
comprised approximately 25%, 33%, and 30% of the increase in total company sales for
1992-1993, 1993-1994, and 1994-1995 respectively.

Schering Plough also experienced a large jump in net income in 1994, the first
full year of US. sales. Specifically, net income from 1991 through 1993 ranged from
$646 million to $731 million. However, in 1994, net income increased by almost $200
million to $922 million (See Figure 3B). While it is impossible to attribute this increase
entirely to Claritin, it does correlate with the substantial increase in Claritin sales during
the same time period.

In addition to demonstrating the impact of a US. drug approval on the operating
performance of a firm, Schering Plough’s experience with Claritin provides insight into
the marketing lag between drug approval and product sales. Conversations with industry
analysts and pharmaceutical sales representatives indicate that most pharmaceutical
companies are able to market a drug that obtains an approved new drug application
within two months. The in-depth analysis of Schering Plough supports the belief that
sales of a new product commence almost immediately after the FDA approval of the new

product.

33

6.3

imp:
the r
from
stocl
May

the p

6.4

firm.
can in
Clam

associ

6.3 Impact of Claritin on market value

Finally, descriptive data suggests that US. Claritin approval also had a substantial
impact on the stock price of Schering Plough Corporation. Figure 3C shows that during
the months of January through March of 1993, Schering Plough’s stock price ranged
from $55 to $60 per share. However, in April, the month Claritin was approved, the
stock price jumped $5 per share, or more than 8%. The stock price continued to rise in
May and June with a June closing price of nearly $70, an increase of more than 15% from

the pre-approval price.

6.4 Summary

The experiences of Schering Plough Corporation, a traditional pharmaceutical
firm, provide an example of the magnitude and speed with which a new drug approval
can impact a pharmaceutical firm. Schering Plough Corporation’s receipt of an NDA for
Claritin clearly provides anecdotal support for all of the hypotheses: NDAS are positively

associated with sales, earnings, stock price, and returns.

34

pm
mm

rem

7.]

film
imm
anan
mdf
R&D
Vat‘ia‘!

know

“her:

Houet
R&De

PTOXy \-

Chapter Seven

EMPIRICAL RESULTS

This chapter provides empirical evidence that supports most of the hypotheses
presented in Chapter 5. The first section of the chapter demonstrates that NDAS are a
reasonable proxy for the economic value created through R&D expenditures. The

remaining sections provide the empirical results from the tests of the hypotheses.

7.1 N DAs: A proxy for knowledge

The hypotheses assume that NDAS are a measure of intangible assets created
through successful R&D activities. Therefore, before testing the hypotheses, it is
important to establish that NDAS are associated with R&D expenditures. Correlation
analysis (table not provided) shows a positive relationship between R&D expenditures
and NDAS. In order to provide a stronger test of the relationship between NDAS and
R&D and to allow for control variables, I perform a regression analysis between the

variables. Pakes and Griliches (1984) provides a model that relates the accumulation of

knowledge to R&D.

k“ = R&D“ + a“ (1)
Where:

kw: the accretion of knowledge for firm j during year t;

R&D“: R&D expenditures by firm j during year t.

However, because k is an unobservable output measure for knowledge gained through
R&D endeavors, Pakes and Griliches use the number of patents granted each year as a

proxy variable for knowledge. Although the number of patents obtained by a firm are

35

available to investors fairly early in the development process, they do not eliminate many
of the uncertainties of assessing R&D success. NDAS, although available later, are a less
uncertain measure of R&D success than patents because they give a company the right to
market a new product. Thus, NDAs are used as a proxy for the accumulation of
knowledge acquired from successful R&D activities.

In addition, it is likely that larger firms have larger R&D budgets. Thus, R&D
could appear to be positively associated with NDAS when, in fact, it is merely a proxy for
firm size. To control for this possibility and to provide assurance that the positive
relationship between R&D and NDAs is not spurious, total assets are added to the Pakes

and Griliches model. This results in the following testable model:

NDA“ = at + B;R&D,-,t + BzAssetsj,t + a“ (2)
Where:
NDA“: the number of New Drug Applications approved by the Food and
Drug Administration for firm j during fiscal year t;
at: year specific intercepts (untabulated);
R&D“: research and development expenditures incurred by firm j during

fiscal year t;
Assets“: total assets for firm j at fiscal year end t.
Because the dependent variable, NDAS, is a count variable, ordinary least squares
regression is not appropriate. Therefore, I test this model using a Poisson regression with
a log-linear link which ensures that the predicted value of NDAS will always be

positive.19 Table 2 shows that for both the traditional and generic pharmaceutical

 

l9For a discussion of regression analysis for count data see Cameron and Trivedi (1998) and Greene (1993).

36

 

companies. R
lortraditioru
generic httttS

assets are not

 

pith NDAs fo
turns is unexp
smaller firms.
relationship b
plausible to h

result of R&l

7.2 Opel
7.2.1 Test

The
f1”HS; the Z

andaneam

.ACCOrdmg 1
"Marlee oflp‘
tallies plOble
2;! rd 61701

em 18 no .

11 Elaine“
were cleam. l;
pmt‘lllt’d "'1 T

companies, R&D has a positive association with NDAS after controlling for firm size.20 2'

For traditional firms the coefficient on R&D is positive and significant at 0.01; and for
generic firms the coefficient on R&D is positive and significant at 0.05. In addition,
assets are not associated with NDAS for the traditional firms but are positively associated
with NDAS for the generic firms. The lack of significance of assets for the traditional
firms is unexpected as it seems that larger firms would tend to have more NDAS than
smaller firms. Despite this, the results provide evidence that there is a positive
relationship between R&D expenditures and the number of NDAS obtained. Thus, it is
plausible to believe that NDAS can be used as a measure of the knowledge obtained as a

result of R&D expenditures.

7.2 Operating models
7.2.1 Tests of hypothesis 1

The real value of the creation of knowledge is the economic benefit it provides to
firms; the Z’s in Figure 2. I test two different models of operating benefits; a sales model

and an earnings model. Relying on models from Behn and Riley (1997) and Lev and

 

20According to Cameron and Trivedi (1998), the data for this regression is overdispersed because the
variance of the dependent variable is more than twice as large as the mean of the dependent variable. This
causes problems similar to hetereoscedasticity in ordinary least squares regression. Therefore, the reported
standard errors are corrected using the method suggested in McCullagh and Nelder (1989).

2|There is no well-established method for deleting influential observations in a Poisson regression other
than evaluating raw residual values. Three of the raw residuals for the traditional pharmaceutical firms
were clearly larger than the other residuals; those three observations were omitted from the results
presented in Table 2. None of the observations for the generic firms were omitted.

37

Sougiannis (1996), I developed the following sales model:

Sales” = or, + BINDAJ-M + BZR&DJ-,t-1 + B3ASSCtSj,t-l + 8);: (3)
Where:

Salesjj revenues, including royalties and licensing fees and excluding non-
operating revenues such as interest, for firm j during year t;

at: year specific intercepts (untabulated);

NDA)“: the number of New Drug Applications approved by the Food and
Drug Administration for firm j during fiscal year t-l;

R&D)“: research and development expenditures incurred by firm j during
fiscal year t-l;

Assets-4-1: total assets for firm j at fiscal year end t-l.

According to Hypothesis 1, the coefficient on NDA should be positive, indicating that the
number of approved new drug applications in a year increases the amount of firm sales in
the following year. As discussed in the Schering Plough case study, drugs obtaining
FDA approval are marketed almost immediately. Therefore, it is possible that no lag is
needed in assessing the relationship between operating performance and drug approvals.
However, I perform the regression analyses with a one year lag of NDAS to ensure that
the new product is on the market for a full year when assessing its impact on sales. R&D
and Assets, both lagged one period, are also expected to be positively associated with
sales. All financial variables, Sales, R&D, and Assets, are included in the model in
logarithmic form to reduce their variance. I do not make any predictions on the year-
specific intercepts but merely include them as additional control variables.

The results for the sales model are shown in Table 3. For traditional

pharmaceutical firms, reported in Panel A, the coefficient on NDAS in the pooled

38

estim;
includ
proxy
also pr
NDA i
signiti
indepe
and Z2
signifit

small 5

traditio
are con:
and sigr
researcl
future 3;
capturin
mCIOgc
annualr;

Wiled tr

flom Iht‘ I6
is: '
‘ "‘ Of or
additional

Joughor

l
\

estimation is positive and significant at 0.01 even though R&D expenditures are also
included in the model.22 This result supports H1 and suggests that new drug approvals
proxy for intangible assets that help to generate future sales. The year-by-year results,
also provided in Table 3 Panel A, provide additional support for H1 as the coefficient on
NDA is positive in nine of the eleven years.23 21 and Z2 statistics are provided to test the
significance of the annual t-statistics. Z1 may be overstated when the t-statistics are not
independent; Z2 corrects for cross-sectional and serial correlation (Barth 1994). Both 21
and Z2 indicate that the annual, cross-sectional t-statistics are significant. The lack of
significance in the annual results is most likely caused by a lack of power as a result of
small sample sizes.

Counter to results found in prior research, the coefficient on R&D for the
traditional firms is negative in the pooled regression. The annual cross-sectional results
are consistent with the pooled results as the coefficient is negative in 10 of the 11 years,
and significantly so in 5 years. The negative coefficient may be driven by the fact that
research and development expenditures in the pharmaceutical industry do not generate
future sales or earnings for several years. Hence, a one-year lag may be insufficient for
capturing any benefit that is directly attributable to R&D. As expected, the coefficient on
the log of Assets is positive and significant in the pooled regression and in all of the
annual regressions indicating that larger firms have more sales. The adjusted R2 for the

pooled traditional model is 0.95.

 

22Following Amir and Benartzi (1998), observations with an R-student value greater than |3| are omitted
from the regression results as these observations are assumed to have a significant influence on the results.
This methodology is followed throughout the paper for the pooled models. In most of the regressions, 1%
to 2% of the observations are omitted. In one case, approximately 6% of the observations are omitted. For
additional discussions of influential observations, see Belsley, Kuh, and Welsch (1980).

23Throughout the paper, annual results are reported when at least ten observations are available.

39

13bit
One]
ofan

{Y ’3
i03:£
~

negaht
fignnu
Onaht
Posent

leteh;f

3365““
Hzandi
asasn]

phamlac

Rather unexpectedly, the coefficient on NDA for the generic firms, reported in
Table 3 Panel B, is negative, although not significant. This result does not support H1.
One potential explanation for this result is that a one-year time difference between receipt
of an NDA and sales is not sufficient for the generic firms. It is probable that it takes
longer for a generic product to generate new sales as the branded product is already
entrenched in the market place. This is especially true as physicians tend to prescribe
products that they are familiar with. To examine this possibility, I ran the sales model for
the generic firms afier substituting NDA” for NDA“. The coefficient on NDA” is not
significant. Alternatively, the lack of significance may be due to the nature of generic
firms. Generic firms earn profits by charging low prices but selling high volumes over
several years. Thus, it is likely that the incremental benefit of one NDA cannot be
detected in the annual sales variable.

Table 3 Panel B also shows that the coefficient on R&D for the generic firms is
negative and significant while the coefficient on Assets is, once again, positive and
significant. The annual results are consistent with the pooled results as the coefficients
on all three variables have the same sign in the four years for which annual data is
presented. In addition, the 21 and 22 statistics show virtually the same significance
levels for all three coefficients.

In summary, for the traditional pharmaceutical companies, the results from the
sales model show that NDAs at H are positively associated with sales at t. This supports
H1 and implies that NDAS can be used as an intangible asset proxy that is useful for
assessing future revenues. H1 is not supported by the results obtained for the generic

pharmaceutical firms as the coefficient on NDA is negative, although not significant, in

40

the pooled regression as well as in the annual regressions. As discussed, this is likely due
to the fact that the incremental benefit of one NDA cannot be detected in the sales
variable.
7.2.2 Tests of hypothesis 2
The earnings model, adapted from Lev and Sougiannis (1996), is:
NIBRDN = a, + BINDAJM + BzR&DJ-,H + B3Assetsj,t-l + a“ (4)
Where:
NIBRDM: income before discontinued operations, extraordinary items,
changes in accounting methods, depreciation and amortization, and
research and development expenditures;

at: year specific intercepts (untabulated);

NDALH: the number of New Drug Applications approved by the Food and
Drug Administration for firm j during fiscal year t-l;

R&D)“: research and development expenditures incurred by firm j during
fiscal year t-l;

Assetst: total assets of firrnj at fiscal year end t-l.

The model predicts that future earnings are a function of tangible and intangible assets.
Depreciation and R&D are added back to operating income since they are write-offs of
the independent variables—tangible and intangible assets. This model differs from the
Lev and Sougiannis model in two ways. First, Lev and Sougiannis do not include non-
financial measures of intangible assets in their model. However, because I am interested
in examining non-financial measures of successful R&D activities, NDAS are added to
the model. Also, Lev and Sougiannis include advertising expenditures as an intangible
asset. Advertising expenditures are definitely important to firms in the pharmaceutical

industry. In fact, according to PhRMA, many firms spend almost as much on advertising

41

and
inta:
finr
the 1
As i
gent

the

pha
nun
in y
is ir

Will

subr

3913'

\‘an'

am
Eli-Diet

“An
C0331

and promotion as they spend on R&D. While it would be ideal to measure this
intangible asset, Compustat only reports advertising expenditures for a few of the sample
firms and this number is rarely reported in the firms’ lO-Ks and annual reports. Thus, in
the price model, advertising expenditures are potentially a correlated omitted variable.
As in the sales model, one lag of NDAS is included in the earnings model as firms
generally are able to market a new drug relatively quickly after obtaining approval from
the FDA.

The results, presented in Table 4 Panel A, support Hypothesis 2 for the traditional
pharmaceutical companies. The earnings model for the traditional firms shows that the
number of new drug applications approved by the FDA in year t-l increases net income
in year t. This is true even though the input measure for knowledge, R&D expenditures,
is included in the model. R&D expenditures and assets are also positively associated
with fiiture earnings. Contrary to the sales model, the coefficient on R&D is positive and
significant in the earnings model. It is possible that this reversal occurs because R&D is
subtracted from net income. More specifically, firms with large R&D budgets may
appear to have larger net income simply because R&D is added back to the dependent
variable. Similar to the sales model and consistent with predictions, assets are positively
associated with NIBRD. The pooled model produces an adjusted R2 of 0.98 with a

sample of 169 observations.” 25

 

2"The large decrease in N from the sales model to the earnings model occurs because negative earnings
observations are dropped when the log is taken.

25An adjusted R2 of 0.98 is relatively high for an accounting research paper. However, this high value is
consistent with the results reported in Lev and Sougiannis (1996).

42

Overall, the annual cross-sectional results, reported for years with at least 10
observations, are consistent with the pooled results. In particular, the coefficient on NDA
is positive in all eleven years and both the Z1 and 22 statistic indicate significance at
0.01. Furthermore, the coefficient on R&D is positive in 8 of 11 years and the coefficient
on Assets is positive and significant in all 11 years.

The results of the earnings model for the generic firms, reported in Table 4 Panel
B, provide weak support for Hypothesis 2 as the coefficient on NDA in the pooled model
is positive and significant at 0.10. The year-by-year results are similar as the coefficient
on NDA is positive and significant in two of the three reported years. However, the mean
coefficient for the annual results is positive but neither the Zl or 22 statistics are
significant. Again, the weak association between the operating measure and NDA is
likely due to the nature of generic pharmaceutical firms. These firms rely on large
volumes with relatively low profit margins. Thus, it is likely that the dependent variable
is not sensitive enough to detect the incremental benefit of one NDA. Meanwhile, R&D
and Assets are both positively associated with future earnings. The model for the generic
firms generates an adjusted R2 of 0.86.

7.2.3 Summary of operating models

The two operating hypotheses, H1 and H2, are supported by the regression results
from the traditional pharmaceutical firms: NDAS are positively associated with both
future sales and future earnings. This implies that NDAS can be used as a proxy for
intangible assets in the pharmaceutical industry. These results support the conclusions

made in Behn and Riley (1997) in which the authors found that non-financial

43

perio

lldtli

ND:

posi

ICS‘

llh

performance measures were useful in predicting operating performance in the airline
industry.

The results from the generic pharmaceutical companies do not support H1 as
NDAs are not associated with future sales. Weak support is obtained for H2 as NDAs are
positively associated with earnings in the pooled model at the 0.10 level. The lack of
results is likely attributed to the fact that generic pharmaceutical firms rely on low-price,
high volume strategies; and, thus the incremental impact of one NDA cannot be detected

in annual operating results.

7.3 Market models

In addition to examining the relationship between NDAS and operating
performance, I test both a stock price model and a market return model. These are
conducted to determine if the number of NDAS, a non-financial measure of R&D
successes, are reflected in investor valuation decisions.
7.3.1 Tests of hypothesis 3

Amir and Lev (1996) examine the usefulness of various GAAP and non-GAAP
measures in predicting the stock price of firms in the wireless communications industry.
They begin their analysis with the following basic model:

Price“ = [30 + BIBVPSJ-j + BzEPSJ-J + a it (5)
Where:

Price“: stock price of firm j at the end of the third month following the
firrn’s fiscal year end;

BVPSN: (total assets - total liabilities)/shares outstanding for firm j at the
fiscal year end t;

44

To

is ir

Likl
fina
Spe
R&
.\D.

pric

m1:

_\
$831?
per Sh;
lilSlrf‘ta

. u

EPSJ-j: earnings per share before extraordinary items, discontinued
operations, and changes in accounting methods, and after adding
back depreciation and amortization and research and development
expenditures for firm j during year t.

To this model, Amir and Lev add two non-financial measures of intangible assets in the
wireless communications industry; POPS, a measure of the population in licensed
metropolitan and rural areas, and PEN, the number of subscribers divided by POPS.26
Likewise, I start with this basic model and add one financial measure and one non-
financial measure of the intangible asset created through R&D expenditures.
Specifically, I include R&D expenditures and the number of NDAS approved each year.
R&D expenditures are included in the model because I am interested in assessing if

NDAS provide incremental information to decision makers when determining the stock

price of pharmaceutical firms. This results in the following model:

Price“ = (11 + BIN-DALI + BzRgLDPSjt + B3EPSJ'J + B4BVPSjt + 8“ (6)
Where:

Price“: stock price of firm j at the end of the third month following the
firm’s fiscal year end;

at: year specific intercepts (untabulated);

NDA“: the number of New Drug Applications approved by the Food and
Drug Administration for firm j during fiscal year t;

R&DPSN: research and development expenditures per share incurred by firm

j during year t;

 

26Barth, Clement, Foster, and Kasznik (1998) also tests intangible assets in a valuation model that is on a
per share basis. Hirschey, Richardson, and Scholz (1998) examines the valuation implication of patents
using the market value of equity scaled by book value of assets as the dependent variable. There is no clear
justification for either method. I chose to use the per share model because the impact of NDAS on a per
share basis is easier to interpret.

45

BPS”: earnings per share before extraordinary items, discontinued
operations, and changes in accounting methods, and after adding
back depreciation and amortization and research and development
expenditures for firm j during year t;

BVPSJ-J: (total assets — total liabilities)/shares outstanding for firm j at fiscal
year end t.

Because a price model examines total firm value at a point in time, it would be ideal to
include the accumulated NDAS across time in the model rather than the approvals in one
particular year. However, it is not clear how long the accumulation period should be as
drugs differ in the amount of time they are used. For example, a truly creative product is
likely to be valuable to a firm for a long period of time. Meanwhile, some products
become obsolete in a few years as more effective and safe replacements are developed.
This same type of issue is present when patent counts are included as independent
variables in a valuation model. I follow the methodology used in Hirschey, Richardson,
and Scholz's 1998 paper in which the authors include annual patent counts as an
independent variable in the valuation model versus the sum of patents over several years.
The results, presented in Table 5, provide support for Hypothesis 3 as NDAS are
positively associated with stock price for both traditional and generic pharmaceutical
firms in the pooled models. For traditional firms, reported in Panel A, the price model
has an adjusted R2 of 0.62 with a sample size of 212.27 NDAS are significant and add
thirty-three cents, or nearly one percent, to the average firm stock price of $41 .27. The
annual cross-sectional models are consistent with the pooled results as the coefficient on

NDA is positive in eleven of the twelve years and significantly so in 1985, 1987 and

 

27The sample size for the price model is larger than that used for the operating models because no lagged
variables are required.

46

1991. Further, the mean coefficient for the annual data is positive and significant. These
results suggest that the number of NDAS obtained by traditional pharmaceutical
companies is viewed as an intangible asset by the stock market and provide support for
Hypothesis 3.

Interestingly, for the traditional pharmaceutical firms, the coefficient on R&DPS
is negative and significant in the pooled model. Also, the annual regressions produce a
negative coefficient in 8 of 12 years as well as a negative mean coefficient. Further, the
21 statistic is negative and significant while the ZZ is negative but not significant. This
result is note worthy because most related literature predicts that R&D and other similar
expenditures (i.e. advertising) will result in an increase in stock price as this type of
investment is actually creating an intangible asset. In fact, Sougiannis (1994) shows that
a one dollar increase in R&D generates, on average, a five dollar increase in total market
value.

There are two possible explanations for the conflicting results found between
pharmaceutical firms and prior research. First, it is possible that R&D expenditures are
incurred relatively early in the pharmaceutical industry in relationship to product releases.
This might cause the market to place little value on the intangible created by the
expenditures. Alternatively, R&D in the earlier research papers might be correlated with
an alternate measure of intangible assets that is omitted from the model. I examined this
possibility for the sample of pharmaceutical firms by running the price regression with
NDAS omitted. The coefficient on R&DPS for the traditional firms remains marginally
negative while the coefficient for the generics is still insignificant. Thus, the first

explanation is more plausible. In general, in the pharmaceutical industry, the number of

47

NDAS appears to be a better measure of the potential marketable products owned by a
firm; hence, R&D is viewed as an expenditure that reduces investors’ equity.

Finally, for the traditional firms, the coefficient on EPS is positive and significant
but the coefficient on BVPS is not significant. This result is surprising but might be an
indicator of the amount of intangible assets that are not included on the balance sheets of
pharmaceutical firms. Amir and Lev (1997) reports findings that are somewhat similar
to the one noted above. In particular, in the wireless communications industry, Amir and
Lev find that BVPS is not significantly related to stock price until non-financial measures
of intangible assets are added to the model. They explain this result by saying that the
financial and non-financial measures compliment each other when assessing a firm’s
value. It is likely that my model still omits several measures of intangible assets,
especially brand loyalty generated by large expenditures on advertising and promotions.
Perhaps if additional non-financial measures of the unreported intangible assets were
included, the results would be consistent with those found by Amir and Lev.

The results from the price model for the generic firms also support H3 as the
coefficient on NDA for the generic firms is positive and significant at 0.01. All six of the
yearly NDA coefficients are positive and three of them are significant. In addition, the
Z-statistics, which summarize the t-value of the annual coefficients, are both significant.
Although this result is consistent with H3, it is somewhat unexpected given that NDAS
were not positively associated with future sales and were only modestly associated with
future earnings. This suggests that, while the incremental benefit of an NDA cannot be
detected in operating performance measures, the market understands the importance of

new drug approvals to the long-run health of generic firms. Thus, for both the traditional

48

and get

recognr

R&DPS

To this r
Spflific;
the Unex
developr
addition

mOdel:

and generic firms, NDAS can be used as a proxy for intangible assets that are not
recognized.

Similar to the results obtained for the traditional firms, the coefficients on
R&DPS and BVPS are not significant while the coefficient on EPS is positively
significant. The pooled model produces an adjusted R2 for the generic firms of 0.39 with
a sample of 120 firms.

7.3.2 Tests of hypothesis 4

I also develop and test a market return model. Similar to Amir and Lev (1996), I
start with the following market return model:

Retumjj = Bo + BJEPSM + BzAEPSj,t + a it (7)
Where:

Retumjaz market return for firm j from the end of the first quarter for year t
to the end of the first quarter for year t+1;

BPS“: net income from continuing operations for firm j during year t,
divided by the number of shares outstanding;

AEPSJ-J: the change in net income per share for firm j from year t-l to year
t, divided by the number of shares outstanding.

To this model, I add a measure for the intangible asset created with R&D expenditures.
Specifically, I include the number of approved new drug applications obtained in a year,
the unexpected new drug applications obtained in a year, a measure of research and
development expenditures, and the change in research and development expenditures. In
addition, time specific intercepts are included in the model. This results in the following
model:

Return“ = (II + BjNDAjJ + BQANDAN + BzR&DPSj,t + B3AR&DPSj't
+ B4EPSJ'J + BSAEPSJ'J + 8.” (8)

49

Where:

Return“:

NDA“:

ANDAJ-j:

R&DPS“:

AR&DPs,-,,:

EPSJ'J:

AEPS‘H:

market return for firm j from the end of the first quarter for year t
to the end of the first quarter for year t+l;

year specific intercepts (untabulated);

the number of New Drug Applications approved by the Food and
Drug Administration for firm j during fiscal year t;

the change in the New Drug Applications approved by the Food
and Drug Administration for firm j from fiscal year t-l to fiscal
year t;

research and development expenditures incurred by firm j during
year t divided by the number of shares outstanding;

the change in research and development expenditures per share for
firm j from year H to year t;

net income from continuing operations before depreciation and
amortization and before research and development expenditures
for firm j during year t, divided by the number of shares
outstanding;

the change in net income per share (from continuing operations
before depreciation and amortization and before research and
development expenditures) for firm j from year t-l to year t,
divided by the number of shares outstanding.

The pooled results for the traditional firms, reported in Table 6 Panel A, do not support

Hypothesis 4. Specifically, neither NDA nor ANDA are significantly associated with

stock returns for the traditional firms. However, the coefficient on ANDA is positive,

although not significant in nine of the eleven reported years and both the Z1 and Z2

statistics for ANDA are positive and significant. Thus, while the pooled model does not

produce significant results, the annual results do provide some support for the idea that

ANDA, the unexpected NDAS, are positively associated with returns.

50

Table 6 Panel A also shows that R&DPS is not associated with firm stock returns
but AR&DPS is negatively associated with stock returns. This result is consistent with
the results reported for the price model and implies that an increase in R&D expenditures
causes the value of the traditional pharmaceutical firms to decrease. Similar to results
documented in prior research, unexpected earnings are positively associated with stock
market returns. Utilizing a sample of 193 firms, the returns model for the traditional
firms produces an adjusted R2 of 0. 1 7.28

For the generic firms, the coefficient on NDA is negatively significant in the
pooled regression. Meanwhile, the results from the yearly analysis show that the
coefficient on NDA is positive in three of the four reported years and that the mean
coefficient is positive although not significant. The lack of significance in the yearly
results is likely due to low power as a result of the small sample sizes. Further, the
negative coefficient on NDA in the pooled model appears to be driven by 1996 as the
coefficient in all other years is positive. The results indicate that there is not a consistent
relationship between ANDA and returns. As in the returns model for the traditional
firms, AR&DPS is negatively associated with stock returns while AEPS is positively
associated with stock returns for the generic firms.

The negative coefficient on AR&DPS for both the traditional and the generic
pharmaceutical firms is counter to the results obtained in prior research which, in general,
show that R&D is positively associated with market performance. As discussed earlier,

one explanation for this result is that R&D is incurred so early in the development cycle

 

28The large negative adjusted st reported in the annual regressions are caused by having a small sample
size in relation to the number of independent variables.

51

of pharmaceutical firms that the market assigns little value to the information provided by
the expense.

The absence of a significant positive relationship between NDAS and returns for
both groups of firms is somewhat surprising given the positive relationship found
between NDAs and stock price in the previous section. The presence of multi-
collinearity between NDAS and ANDA could produce inefficiencies in the model and
cause NDAS to be insignificant. However, correlation analysis shows that NDA and
ANDA are only weakly correlated. Therefore, inefficiency as a result of multi-
collinearity does not appear to be the cause of the insignificance.

An alternate reason for the insignificance of NDA in the return model is that the
release of product approval information and the return data are not properly matched.
The results reported in Table 6 use fiscal year NDAS and financial data while the returns
start with the first quarter of year t and are calculated through the first quarter of year t+1.
Most prior research uses first quarter returns to ensure that all of the firms have released
their financial data and that the market has had time to incorporate the data. However,
the NDAS are available immediately upon approval by the FDA. Thus, it might be more
reasonable to match returns with the fiscal year to investigate the impact of NDAS on
market returns. Hence, I calculated market returns during the fiscal year and ran the
regression analysis again. The results are not materially different from those reported in
Table 6. Specifically, the coefficients on NDA and ANDA are not associated with returns
for the traditional firms. For the generic firms, the coefficient on NDA is negative and
significant while the coefficient on ANDA is also negative, although not significant. The

sign and significance of the other coefficients are virtually unchanged.

52

The results from the returns model do not support Hypothesis 4. In particular,
while NDAS are positively associated with stock price for the both traditional and generic
pharmaceutical firms, they are not positively associated with stock returns. Most
surprising is the fact that the coefficient on ANDA is negatively associated with returns
for the generic firms, albeit at 0.10. Hirschey, Richardson, and Scholz (1998) reports
similar results for patent data. In their paper, the number of patents is positively
associated with stock price but is not significantly associated with stock returns. Perhaps
both of the results have the same problem--using a relatively long window for predicting
returns for a specific event. It is likely that numerous other factors affect stock returns
over a one-year time period and may overshadow the impact of new drug approvals.
7.3.3 Summary of market models

The results for the market models are mixed. The results from the price model for
both the traditional and generic pharmaceutical firms support Hypothesis 3 and indicate
that the number of new drug approvals, a non-financial measure of R&D success, is
helpful in assessing firm value. In addition, the insignificance of BVPS suggests that
current balance sheets provide little information when assessing the value of firms in the
pharmaceutical industry. I expect that these results would also hold in other industries
that are highly dependent on intangible assets that are not measured on a firm’s balance
sheet. In fact, Amir and Lev (1997) shows that neither BVPS nor EPS are associated
with stock price in the wireless communications industry when solely included in a price
model. Therefore, finding alternate measures of key intangible assets may be critical for

assessing value in industries which are dependent on these unrecognized assets.

53

Despite the fact that NDAs are positively associated with stock price, they are not
positively associated with stock returns. Hirschey, Richardson, and Scholz (1998) finds
similar results for patent data. These results can most likely be attributed to the fact that a
relatively long window, one year, is used for evaluating a market reaction to an event.
Future research that examines a short-window surrounding the receipt of an NDA might

prove fruitful.

7.4 Sensitivity analysis: Hl-H3

As discussed earlier, an alternate measure of the knowledge created through R&D
expenditures is the number of new patents obtained by a firm. I discussed several reasons
why this measure might not be the ideal choice for economically valuable knowledge in
the pharmaceutical industry. However, to provide additional support for the results
obtained previously, I ran the regression models for the traditional firms including patent
counts as an independent variable.” 30 The first column of Tables 7, 8, and 9 provides
the results for the sales, earnings, and price models afler including one year of patent
data.

In the Sales model, the coefficient on NDA is positive and significant even though

patents at t-l are also included in the model. As in the original sales model, R&D is

 

29 I collected the patent data from the US. Patent Office (1999). Because of the way the database works,
the patent data is gathered on an annual basis (i.e. it is not based on a firm’s fiscal year end). Thus, the
reported results using patent data contain noise to the extent that firms have fiscal year ends.

30Patent counts are only added to the models for traditional firms as generic firms, in general, do not apply
for patents. To make sure that this assumption is valid, I searched the US. patent database for 6 of the 16
largest generic pharmaceutical companies. For the twelve year period from 1985 to 1996, the 6 firms (72
firm-years) obtained a total of 3 patents between them. Meanwhile, the mean number of patents for the
traditional pharmaceutical companies is 51 and the mode is 38.

54

negatively associated with sales while assets at t-l are positively associated with sales.
Utilizing 185 observations, the model generates an adjusted R2 of 0.96.

It is also possible that because of the length of time involved in the development
process that additional lagged variables should be included for R&D expenditures and
patent data. Therefore, I add five years of data for both patent counts and R&D
expenditures in the sales model. The results are essentially the same as those with one
year of R&D and patent data. The only notable change is that the coefficient on patents
at H changes from positive and significant to negative and insignificant. This result is
likely caused by the high multi-collinearity in the model. Specifically, the five years of
patent data are correlated between 0.86 and 0.96, causing the model to be inefficient.

The results for the earnings model including the patent data continue to support
H2. The coefficient on NDA remains positive and significant even though the number of
patents received during year t-l is included in the model. The coefficients on patents,
R&D, and Assets at t-l are also positive and significant. When five years of patent and
R&D data are included in the model, the coefficient on NDA is still positively associated
with earnings. As in the sales model, none of the patent variables are significant. Similar
to results found in Sougiannis (1996) the R&D coefficients vary from positively to
negatively significant. This implies that in some years the cost of R&D outweighs its
benefits.

The results for both the sales and earnings models that include patent data support
the conclusions made previously. Specifically, each NDA received in year t-l increases
sales and earnings in year t. And, while patents at H are positively associated with both

sales and earnings, they do not provide the same information as that provided by NDAS.

55

Thus, NDA information is useful above and beyond that of R&D and patent data in
assessing future sales and earnings.

The results for the price model, including patents, are also consistent with the
original price model. The coefficient on NDAS is 0.28 and is significant at 0.01. In
addition, the coefficient on patents is positively associated with price. Similar to the
results provided in Table 5, the coefficient on R&DPS is negatively associated with price.
EPS is positively associated with stock price at 0.01 while BVPS is marginally associated
with stock price.

Including five years of patent and R&D data in the model generates somewhat
different results. First, as in the model with one year of patent data, NDAs remain
positively associated with stock price at 0.01. Meanwhile, the significance level for
BVPS increases from 0.10 when one year of patent data is included to 0.01 when five
years of patent and R&D data are included. The BVPS result is interesting because
BVPS was not associated with price in the original model and was only marginally
associated with price when one year of R&D and patent data are included. Meanwhile,
the significance level of EPS decreases from 0.01, when only one year of patent and
R&D data are included, to 0.10 when five years of data are included. These shifts in
levels of significance are difficult to explain.

Another interesting result occurs as the coefficient on patents at time t is negative
and significant when 5 years of R&D and patent data are included in the model. This
result is likely due to multi-collinearity in the patent data. One final change is that EPS is
no longer positively associated with price when the model is expanded to include five

years of R&D and patent data. I have no explanation for this result.

56

Overall, the results shown in Tables 7, 8, and 9 suggest that NDAS provide
information to investors that is not provided by patent counts. Perhaps NDAS and patents
provide two pieces of information in a puzzle. For example, the patent data provides
early information on R&D progress while the NDAS provide data about products that will

be marketed in the near future.3 I

7.5 Successful developer models

As discussed in the hypothesis development section, it is likely that some firms
are better at developing new drugs for their pipelines than other firms. The R&D of the
efficient developers should result in larger increases in sales and earnings and therefore
should be valued more highly by the stock market. This idea is tested by partitioning the
firms into successful and non-successfirl groups based on the number of NDAs obtained.
To make the division, I first computed the mean number of NDAS obtained in a year by
each firm. I then calculated the median of the firm means for both the traditional and
generic group of firms. Any firm in which the mean NDAs is greater than the respective
median is classified as a successful firm. Both the price and return model are adjusted
using the following methodology. First, NDA is replaced with a dichotomous variable
that is set to 1 if the firm is deemed to be successful and 0 otherwise. Then, an
interaction between successful firms and R&D expenditures is included. Hypotheses 5

and 6 predict that the coefficient on the interaction term will be positive.

 

3 ' I did not run the patent sensitivity analysis on the return model since Hypothesis 4 is not supported.

57

7.5.1 Tests of hypothesis 5

To test H5, I modified the price model to include the successful firm dichotomous
variable and the interaction between success and R&D expenditures. The results from
this model are reported in Table 10. For the traditional pharmaceutical companies, the
Success coefficient is positive and significant. This result provides further support for H3
and suggests that there are intangible assets that are not being recorded on the balance
sheets of pharmaceutical companies. The coefficient on R&DPS is negative although not
significant.

H5 asserts that the coefficient on the interaction between Success and R&DPS will
be positive as the R&D of successful firms should be valued more highly than the R&D
of non-successfirl firms. The pooled results for the traditional firms support this
assertion, as the coefficient on the interaction is positive and significant. In fact, the
results indicate that the R&D of non-successful firms is not valued by the stock market at
all as the coefficient on R&DPS is not significant. As expected, EPS and BVPS are both
positively associated with stock price.

The annual regression results reveal an interesting phenomenon. In particular,
there seems to be a change in the way the stock market values pharmaceutical firms after
1991. From 1985 to 1990, the coefficient on Success is positive in each of the six years
with a mean coefficient for this period of 27.38. During the last six years, however, the
magnitude of the coefficients decreases generating a mean of only 11.96.

The coefficients on R&DPS also change from the first six years to the last six
years. From 1985 to 1990, the coefficients on R&DPS are always positive and generate a

mean coefficient of 20.53. Beginning in 1991, five of the six coefficients are negative

58

and produce a mean coefficient of —4.97. Thus, from 1985 to 1990, the summarized
annual coefficients on R&DPS are positive and significant with a ZZ statistic of 3.20
while from 1991 to 1996, the summarized annual coefficients on R&DPS are negative
and significant with a Z2 statistic of —2.21.

Finally, the annual coefficients on the interaction between Success and R&DPS
also change from the first six years to the last six years. From 1985 to 1990, the
coefficient is negative in all six years and significantly so in 3 of those years. The annual
data produces a mean coefficient of —19.29 and a Z2 statistic of —3.36, which is
significant at 0.01. However, from 1991 to 1996, the coefficient on the interaction is
positive in five of the six years with a mean coefficient of 4.94 that is significant at 0.01.

These dramatic shifts in coefficient values from the first six years to the second
six years are very intriguing. During the first six years, the magnitude of the coefficient
on Success is very large. However, from 1991 to 1996, the magnitude on this coefficient
starts to decline and the significance level begins to drop. In particular, during the first
six years, three of the annual coefficients are positively associated with stock price while
only one is positively associated with stock price during the second six years.

At the same time, the ability to value R&D expenditures seems to become more
refined. From 1985 to 1990, R&D, on average, is positively valued by the market.
However, the R&D of successful firms appears to be negatively associated with price.
Counter to the early results, from 1991 to 1996, the R&D of successful firms is valued
more positively by the market than the R&D of non-successful firms. In particular, the
results show that the coefficient on the interaction between Success and R&DPS is

positive in five of the six years and significant in three of those five years. During this

59

same time period, the R&D of non-successful firms is negatively valued by the market.
These results seem to suggest either that investors learn which firms are successful as
time passes or that investors become better at interpreting the data that is available to
them. It is likely that both of these explanations contribute to the change in the
relationship between the dependent and independent variables.

It is possible that the change in results is caused by the method used to segregate
the successfirl and non-successful firms. In particular, I separate the firms into successful
and non-successful groups based on the mean number of NDAS obtained over the entire
sample period. Investors do not have access to all of this data when they are valuing the
firms. For example, I use information available in 1996 to classify firms in 1985.
Investors only have access to data that is available in 1985. Thus, it is possible that as
investors gain more knowledge, their distinction between successful and non-successfiil
firms more closely matches mine and drives the results.

For the generic pharmaceutical firms, Success is positively associated with stock
price. This provides additional support for the results reported in Table 5 and suggests
that NDAS are a measure of intangible assets. Contrary to the results found for the
traditional firms, R&DPS is positively associated with stock price while the interaction
variable is negatively associated with stock price. This result is difficult to interpret. As
before, EPS is positively associated with price while BVPS is not. The results from the
five reported annual regressions are consistent with the pooled results.

Overall, the results from the successful price model provide mixed support for H5.
For the traditional firms, the coefficient on R&DPS for non-successful firms is not

significant. However, the interaction between R&DPS and Success is postively

6O

associated with price. This implies that the R&D of successful firms is valued by the
market and that the R&D of non-successful firms is not. The results from the generic
firms, meanwhile, do not support H5 as the coefficient on the interaction variable is
negative and significant. This result is surprising.

The results from the successful price model also provide additional support for H3
as the success intercept is positive and significant for both the traditional and generic
firms. The magnitude of the coefficient on Success suggests that successful firms have
significant intangible assets that are not currently being reported on their balance sheets.
7.5.2 Tests of hypothesis 6

The results from the successful returns model are reported in Table 11. The
pooled results for the traditional and generic firms do not support H6. In particular, the
coefficient on the interaction between Success and R&DPS is not significant for either
group of firms. The year-by-year analyses provide similar results. For the traditional
firms, the coefficient on Success*R&DPS is positive in six of the eleven years and
significantly so in one of those six. The coefficient is significantly negative in four of the
remaining five years. These conflicting results make it difficult to draw conclusions
about the model. Further, for the generic firms, the coefficient is not associated with
returns in any of the four reported years. The coefficient on Success*AR&DPS is not
significant for either the traditional or generic firms. Similar to the results reported in
Table 6, the change in EPS is positively associated with returns for both the traditional
and generic firms.

The results reported in Table 11 do not support H6 as the R&D of successful firms

is not more positively associated with market returns than the R&D of non-successful

61

firms. This result is not surprising given the results reported for the returns model in

Table 6.

7.6 Summary

This chapter presents the empirical results from the tests of the six hypotheses. In
support of the operating models, NDAS are positively associated with one-year ahead
sales for the traditional firms and with one-year ahead earnings for both the traditional
and generic firms. In addition, NDAS are positively associated with stock price for both
groups of firms. These results suggest that NDAS can be used as a proxy for intangible
assets not recorded on the balance sheets of pharmaceutical firms. Neither NDAS nor the
unexpected NDAS is associated with market returns for either group of firms. As
discussed, this result is likely due to the fact that a relatively long return window was
used, one year, to assess the market response to an event. Future research might benefit
by an event study that assesses market reaction around the date the NDA is received.

The results from the successful product developer hypotheses are mixed. The
empirical results for the traditional pharmaceutical firms support the price model, H5, as
the R&D of successfirl firms is positively associated with firm market value while the
R&D of non-successfirl firm is not. In addition, being classified as a successful firm
increases stock price; this provides additional support for H3 and suggests that there are
intangible assets that are not recorded on the balance sheet of pharmaceutical firms. The
results for the generic firms do not support H5 as the R&D of successful firms is
negatively associated with stock price. However, similar to the traditional firms,

additional support is provided for H3 as the dichotomous success variable is positively

62

associated with price indicating that the non-recognized intangible assets of successful
firms are valued more than the intangible assets of non-successful firms. Finally, the
regression results do not support H6 for either group of firms as the R&D of successful

firms does not generate larger returns than the R&D of non-successful firms.

63

Chapter Eight

CONCLUSION

This paper examines a non-financial measure for the intangible asset created with
R&D expenditures in the pharmaceutical industry, namely the drug pipeline. After
confirming that NDAS are a valid measure of the accumulation of knowledge, I examined
if NDAS are associated with the following measures of economic performance: future

sales, future earnings, current stock price, and market returns.

8.1 Review of results
8.1.1 Summary of the results for traditional pharmaceutical companies

To summarize, I will first highlight the results from the traditional pharmaceutical
companies. Hypothesis 1 asserts that the number of approved new drug applications will
be positively associated with future sales. This hypothesis is supported by the empirical
results as the coefficient on NDAS is positive and significant in the sales model.

Further, the regression analysis for the earnings model shows that each NDA
obtained by the traditional companies has a positive impact on one period ahead
operating earnings. Thus, H2 is supported. The anecdotal evidence discussed in the case
study provides additional support for H2. In fact, Schering Plough’s net income increased
by almost $200 million the year after the FDA approval of Claritin. These results are
consistent with those found in Fomell, Ittner, and Larcker (1996) and Behn and Riley
(1996) which found that non-financial indicators can provide information that is useful in

assessing future operating performance.

Because the number of NDAs is positively associated with future sales and
earnings, they should also be valued by the stock market. Hypothesis 3 predicts that the
number of NDAS will be positively associated with the current stock price. The
empirical results for the traditional pharmaceutical companies are consistent with this
hypothesis. In particular, each NDA obtained during year t increases the three month out
stock price by approximately one percent. The US. approval of Claritin by Schering
Plough further demonstrates how large the impact on price can be. In the month that
Claritin was approved, the stock price of Schering Plough increased from approximately
$60 per share to approximately $65 per share, or more than 8%. This result suggests that
there are substantial intangible assets that are not recognized on the balance sheets of
traditional pharmaceutical companies. It is therefore important that investors consider
alternate measures of the assets that are not recorded.

Although there is a relationship between NDAS and stock price, there is no
apparent relationship between NDAS and market returns. The lack of an association
between NDAS and returns is likely a result of using a long window to evaluate investor
reaction to a particular event. Hirschey, Richardson, and Scholz (1998) obtained similar
results when assessing the relationship between patent data and market performance
measures. In particular, they found that patent information was positively associated with
stock price but not with annual stock returns. In the future, it would be beneficial to
conduct an event study surrounding the release of drug approval information.

To check the robustness of the NDA results, I included the number of patents
obtained by the traditional firms in the sales, earnings, and price models to ensure that

NDAS provide information that is not available in patent counts. In all three models,

65

NDAS are positively associated with the dependent variables after controlling for patents
and R&D. As a final verification of the models, I included five years of R&D and patent
data in the models. Once again, NDAS are positively associated with future sales, firture
earnings, and current stock price. This implies that approved new drug applications
provide information useful for assessing future operating performance and current stock
price that is not available in R&D and patent data.

Finally, analyses were conducted to determine if the research and development
expenditures of successful firms are more highly valued than the research and
development of non-successful firms. Successful firms are identified as those with a
relatively higher level of NDAS per year. The coefficient on Success is positive. This
result supports the earlier conclusions that NDAs are an indicator of unrecognized
intangible assets. In addition, the interaction between Success and R&DPS is positively
associated with stock price indicating that the R&D of successfiil product developers is
valued more by investors than the R&D of non-successfiil producers. The return model
exhibits no differences between successful and non-successful firms.

Overall, the results for the traditional pharmaceutical companies provide support
for the hypotheses. NDAs are positively associated with future earnings, future sales, and
current stock price. In addition, a dichotomous variable that distinguishes successful
from non-successful product developers is positively associated with stock price as is the
interaction between the dichotomous variable and R&D. These results imply that NDAS
are a proxy for economically valuable knowledge in the pharmaceutical industry. The
results also suggests that investors value the R&D of successful firms more than the R&D

of non-successful firms.

66

8.1.2 Summary of the results for generic pharmaceutical companies

The results for the generic pharmaceutical firms are mixed. First, Hypotheses 1 is
not supported as NDAS are not positively associated with future sales. This result might
be attributed to the fact that generic firms rely on low-price, high volume market tactics.
Therefore, it is possible that the incremental benefit to receiving an NDA cannot be
detected in sales for a single period. NDAS are marginally associated with future
operating earnings.

Despite the weak operating results, NDAS are positively associated with stock
price for the generic firms. This suggests that NDAS can be used as a proxy for the
intangible asset created with investments in R&D. As with the traditional firms, there is
no relationship between NDAS and returns.

Finally, the results from the generic pharmaceutical firms do not support the
successful product producer hypotheses. While being classified as a successful product
producer is positively associated with stock price, the R&D of successful firms is not
positively associated with price. More surprising, the R&D of successful firms is
negatively associated with stock price. This result is difficult to interpret. As with the
traditional firms, there is no relationship between successful firms and stock market

returns.

8.2 Summary
The analyses confirm the results found in other research (Amir and Lev 1996;
F omell, Ittner, and Larcker 1996; Behn and Riley 1997; Barth, Clement, Foster, and

Kasznik 1998; and Hirschey, Richardson, and Scholz 1998). In particular, non-financial

67

measures of economic value can be used to provide additional information in firm
valuation decisions. More specifically, non-financial measures of R&D success provide
information not contained in R&D expenditures. This documents that providing non-
financial information (i.e. NDAS) to investors and analysts enhances their ability to
predict operating and market performance of firms in the pharmaceutical industry.
Measuring intangible assets will continue to be an important focus of standard
setters, decision makers, and investors. In this light, several streams of future research
might prove beneficial. First, it is interesting that the results for the traditional and
generic pharmaceutical companies are not the same. Future research could strive to
identify an alternate measure of intangible assets for the generic firms. In addition,
advertising and direct promotion to physicians and hospitals creates another important
intangible asset to pharmaceutical firms. A study that incorporated a value for this
intangible would enhance the understanding of this industry. Further, an event study
focusing on the time surrounding the release of new drug approvals by the Food and
Drug Administration might provide additional insight into the valuation of successful
R&D endeavors. Also, research that investigates methodologies for assigning value to

new products would make strides toward valuing intangible assets.

68

LIST OF REFERENCES

69

LIST OF REFERENCES

Aboody, D. and B. Lev. 1998. “The value relevance of intangibles: The case of software
capitalization.” Journal of Accounting Review 36 (Supplement): 161-191.

AICPA Special Committee on Financial Reporting. 1994. “Improving business
reporting—A customer focus.” (New York; American Institute of Certified Public
Accountants).

Amir E. and S. Benartzi. 1998. "The expected rate of return on pension firnds and asset
allocation as predictors of portfolio performance." The Accounting Review 73
(July): 335-352.

Amir, E. and B. Lev. 1996. “Value relevance of nonfinancial information: The wireless
communications industry.” Journal of Accounting & Economics 22 (August-
December): 3-30.

Barth, ME. 1994. “Fair value accounting: Evidence from investment securities and the
market valuation of banks.” The Accounting Review 69 (January): 1-25.

Barth, M.E., M.B. Clement, G. Foster, and R. Kasznik, 1998. "Brand values and capital
market valuation." Working paper, Stanford University.

Beary, J .D. 1997. “The drug development and approval process.” [Online] Available:
hnp://www.phrrnaorglchargs/approval.htrnl [September 8, 1997].

Behn, B. K., and R. A. Riley, Jr. 1997. “Using non-financial information to predict financial
performance: The case of the US. airline industry.” Working paper, The University
of Tennessee. (January).

Belsley, D., E. Kuh, and R Welsch. 1980. Regression diagnostics, identifying influential
data and sources of collinearity. Wiley, New York.

Bound, J., C. Cummins, Z. Griliches, B.G. Hall, and A. Jaffe. (1984) “Who does R&D and
who Patents?” in Griliches: 21-54.

Bouwman, M.J., P. F rishkoff, and PA. Frishkoff. 1995. ”The relevance of GAAP-based
information: A case study exploring some uses and limitations.” Accounting

Horizons 9 (December): 22-47.

Cameron, A. C ., and P. K. Trivedi. 1998. Regression analysis of count data. Cambridge
University Press.

70

Chambers, D., R. Jennings, and RB. Thompson, 11. 1998. “Evidence on the usefulness of
capitalizing and amortizing research and development costs.” Working paper,
University of Texas at Austin.

Davidow, William. 1996. “Why profits don’t matter.” Forbes ASAP. (April): 24.

DiMasi, J .A., R.W. Hansen, H.G. Grabowski, and L. Lasagna. 1991. “Cost of innovation in
the pharmaceutical industry.” Journal of Health Economics 10: 107-142.

Dranove, D. and D. Meltzer. 1994. “Do important drugs reach the market sooner?” Rand
Journal of Economics 25 (Autumn): 402-423.

Elliott, R.K. 1992. ”The third wave breaks on the shores of accounting.” Accounting
Horizons (June): 61-85.

Financial Accounting Standards Board (FASB). 1974. “Accounting for research and
development costs.” Statement of Financial Accounting Standards No.2.

F omell, C., CD. Ittner, and DP. Larcker. 1996. “The valuation consequences of customer
satisfaction.” Working paper, Wharton School, University of Pennsylvania.

Fox, Justin. 1996. “Searching for nonfiction in financial statements.” Fortune. (December
23): 39-40.

Greene, W. H. 1993. Econometric Analysis 2nd edition. Macmillan Publishing Company.

Hall, B.H., Z. Griliches, J .A. Hausman. 1986. “Patents and R&D: Is there as lag?”
International Economic Review (June): 265-283.

Hirschey, M. 1982. “Intangible capital aspects of advertising and R&D expenditures.” The
Journal of Industrial Economics. (June): 375-390.

Hirschey, M., VJ. Richardson, S. Scholz. 1998. “Value relevance of nonfinancial
information: The case of patent data.” Working paper, University of Kansas.

Hirschey, M. and J .J . Weygandt. 1985. “Amortization policy for advertising and research
and development expenditures.” Journal of Accounting Research (Spring): 326-
335.

International Accounting Standards Committee (IASC). 1998. Standard No. 38: Intangible
Assets, London: IASC.

Lev, B. and T. Sougiannis. 1996. “The capitalization, amortization, and value-relevance of
R&D.” Journal of Accounting and Economics 21 : 107-138.

71

McCullagh, P., and J. A. Nelder. 1989. Generalized linear models, edition 2, London,
Chapman and Hall.

Myers, Randy. 1996. “Getting a grip on intangibles.” CF 0 (September): 49-54.
North American BioTechnology Directory. 1995/1996 edition. IEI Publishing Division.

Ohlson, James A. 1995. “Earnings, book values, and dividends in equity valuation.”
Contemporary Accounting Research 11 (Spring): 661-687.

Pharmaceutical Researchers and Manufacturers Association (PhRMA). 1997. “Industry

Profile” [Online]. Available: hflpifl wwwpharma.org/facts/industry/index.html
[September 8, 1997].

Pakes, A. and Z. Griliches. 1984. “Patents and R&D and the firm level: A first look.” In

R&D, Patents, and Productivity. Zvi Griliches (editor), Chicago: Chicago University
Press.

Schweitzer, SO. 1997. Pharmaceutical Economics and Policy. New York: Oxford
University Press.

Sougiannis, T. 1994. "The accounting based valuation of corporate R&D.” The Accounting
Review 69: 44-68.

Standard & Poor’s Corporation Records. 1998. (New York: Standard & Poor’s).

Toole, AA. 1997. “The impact of federally firnded basic research on industrial innovation:
Evidence from the pharmaceutical industry.” Working paper, Laurits R. Christensen
Associates.

US. Patent Office. 1999. [Online]. Available: http://patents/usptg.gov/indexhtml [March 1999].

Wallrnan, S.M.H. 1995. “The firture of accounting and disclosure in an evolving world: The
need for dramatic change.” Accounting Horizons 9 (September): 81-91.

Wallrnan, S.M.H. 1996. “The future of accounting and financial reporting part H: The
colorized approach.” Accounting Horizons 10 (June): 138—148.

Ward’s Business Directory of US. Private and Public Companies. 1998. (Detroit, MI: Gale
Publishing).

White, H. 1980. "A heteroskedasticity-consistent covariance matrix estimator and a direct
test for heteroskedasticity. Econometrica 48: 817-838.

72

FIGURES AND TABLES

73

Figure 1

Pharmaceutical R&D Timeline

 

Market Analysis
(pre R&D activities)

 

 

 

Mean Time (months)

 

Economic
potential

 

 

 

 

 

 

  
  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

42 6 Basic Research,
' Computer Modeling,
and Animal Studies
Investigational New
Drug Application
15.5
Clinical Phase 1:
Testing on healthy
volunteers
24.3
Clinical Phase 11:
Testing on
individuals with
condition
36.0
Clinical Phase III:
Hospital and
Pre-approval total 119.4 outpatient testing
Approval &3
Total 49.7

 
  

 
    

New Drug
Application

Diagram adopted from Schweitzer (1997), Figure 1.3.

74

Figure 2

The Knowledge Production Function

 

 

 

 

 

R&D
Where:
R&D: research and development expenditures
S: measure of R&D success, econorrrics papers usually use patents granted
k: additions to economically valuable knowledge
Z’s: indicators of expected or realized benefits from invention
X’s: other observed variables influencing the 2’s

Diagram adopted from Pakes and Griliches (1984), Figure 4.

75

Figure 3

Schering Plough Corporation

 

 

3A: Change in sales (000's)

 

800
600 .- "
400 H 7 _7
200 ’ UTotal
IClaritin
0 Z v _
91-92 92-93 93-94 94-95
FDA
Approval

 

 

 

38: Net income (000's)

 

1991 1992 1993 1994 1995

FDA
Approval

 

76

 

 

Figure 3 (continued)

 

 

75
70

65 ._
60 .
55 .
50 .

3C: Stock Price

A i _ __ ’ .’__;. _
f z ,0- — — - '9' ,
/ /
.. ‘ ‘ ., ’ a * _ _
T ‘0' ’ 4 if -
Jan Feb Mar April May June
FDA Approval

 

77

 

Table 1

Descriptive Statistics8
Panel A: (11 )

Variable Mean Deviation ' Median
Assets 1
1 1 , l
1

Panel B: Traditional Firms (23 firms, 220
Assets 5 68 1.
4,564
44

6.
.1

Panel C: Firms (80 firms, 353
Assets .4
-64

0.0

Panel D: Generic Firms (16 firms, 122
Assets 1 7 1 1 .
107 1
1

7
l .

 

Where:

Assets total assets at the fiscal year end;

Equity total equity at the fiscal year end;

Sales revenues, including royalties and licensing fees and excluding non-operating
revenues such as interest;

R&D research and development expenditures incurred during the year;

NI net income before discontinued operations, extraordinary items, and changes in
accounting methods;

Price stock price at the end of the first quarter following the fiscal year end;

NDA the number of New Drug Applications approved during the fiscal year for each

firm by the Food and Drug Administration.

aAccounting variables are in millions of dollar.

78

Table 2

NDA Modela’b

 

 

 

 

 

 

 

 

 

Coefficient
(Standard Error)
[Chi-Square]c
R&D” Assets).J N
Traditional 2.0630 -0.0105
Firms (0.4956) (0.0382) 217
[17.33]3 [0.08]
Generic Finns 27.4562 1.5448
(11.7039) (0.8378) 122
[5.50]2 [3.40]'
Where:
NDA“: the number of New Drug Applications approved by the Food and Drug
Administration for firm j during fiscal year t;
at: year specific indicator intercepts (untabulated);
R&D“: research and development expenditures incurred by firm j during year t
Assets“: total assets for firm j at fiscal year end t.

1This model is fitted using a Poisson regression with a log-linear relationship.
”R&Dj,t and Assets];t are included in the model in billions of dollars.
cl, 2, and 3 indicate significance at the 0.10, 0.05, and 0.01 levels respectively.

79

 

Table 3

Sales Modela

Sales“ : a! + B1 NDAj.r-l + BZR&Dj.r-l + B3ASSCISJ-H + sj.r

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Coefficient
(Standard Error)
[adjusted r or Z]b'°"’
Panel A: Traditional Firms
NDAj.t.1 R&D,“ Assetsj_..[ Adj. R‘ N
Pooled with 0.0096 0.0800 1.1642
year (0.0030) (0.0586) (0.0631) 0.95 185
intercepts [3.20]3 [-1.37]' [18.45]3
0.0134 -0.8690 2.4551
1986 (0.0246) (0.3454) (0.4400) 0.96 14
[0.54] {-2.52]2 [5.58]3
0.0000 0.3611 1.5612
1987 (0.0045) (0.0788) (0.1156) 0.99 14
[0.00] [.458]3 [13.51]3
0.0004 0.4219 1.7033
1988 (0.0079) (0.1493) (0.2146) 0.97 14
[0.05] [.283]3 [7.94]3
0.0037 0.3987 1.6434
1989 (0.0057) (0.1674) (0.2244) 0.96 13
[0.65] [.238]2 [7.32]3
0.0083 0.1121 1.2085
1990 (0.0146) (0.0937) (0.0916) 0.98 16
[0.57] {-1.20] [13.19]3
0.0115 0.0293 0.9770
1991 (0.0443) (0.1680) (0.1770) 0.95 16
[0.26] [0.17] [5.52]3
0.0321 0.0957 1.0581
1992 (0.0202) (0.1212) (0.1207) 0.97 19
[1 .59]1 [0.79] [8.77]3
0.0083 0.2531 1.4815
1993 (0.0323) (0.1759) (0.2022) 0.96 20
[0.26] {-1.44]' [7.33]3
0.0190 0.1541 1.2861
1994 (0.0198) (0.1918) (0.2010) 0.95 19
[0.96] [0.80] [6.40]3
0.0205 0.0506 1.0777
1995 (0.0203) (0.1922) (0.1903) 0.95 20
[1.01] [0.26] [5.66]3
0.0183 0.1133 0.8704
1996 (0.0208) (0.1678) (0.1708) 0.96 20
[0.88] [0.68] [5.10]3
Across Years 0.0099 -0.2393 1.3929
(0.0124) (0.2680) (0.4502)
21 [1.62]2 [.457]3 [24.35]3
22 [2.76]3 [—3.15]3 [8.41]3

 

 

 

 

 

 

80

 

Where:

Table 3 (continued)

 

 

 

 

 

 

 

 

 

Panel B: Generic Firms
NDA”. R&D,“ Assets}.-. Adj. R" N
Pooled with 0.0045 0.3546 1.2731
year (0.0054) (0.0528) (0.0528) 0.93 94
intercepts [0.83] [.672]3 [24.1 1]3
0.0016 0.3282 1.2342
1993 (0.0535) (0.1942) (0.1749) 0.88 12
[0.03] [-1.69]' [7.06]3
0.0091 0.3017 1.3081
1994 (0.0169) (0.1241) (0.1235) 0.96 13
[0.54] [.243]2 [10.59]
0.0022 0.3191 1.3705
1995 (0.0162) (0.1010) (0.1091) 0.97 14
[0.14] [.316]3 [12.56]3
0.0359 0. 2785 1.2671
1996 (0.0234) (0.11252) (0.1151) 0.97 13
{-1.53]1 [ 2 48] [1 101]3
Across Years -0.0122 -0.3069 1.2950
(0.0162) (0.0219) (0. 058 7)
21 [-103] [.449]3 [18.97]
22 [-1.4z]' [-704] 3 [7.67]

 

 

 

 

 

 

 

Sales“

01,:

NDAj.l-l:

R&DLH:
Assetsj,,-.:

revenues, including royalties and licensing fees and excluding non-operating
revenues such as interest, for firm j during year t;

year specific indicator intercepts (untabulated);

the number of New Drug Applications approved by the Food and Drug
Administration for firm j during fiscal year t-l;

research and development expenditures incurred by firm j during fiscal year t-l;
total assets of firm j at fiscal year end t-l .

aAll financial variables, Sales, R&D, and Assets, are in millions of dollars. Further, they are included in the

model 1n logarithmic form.

t’l, 2, and 3 indicate significance at the. 10, .05 and .01 levels respectively for a one- -tailed t-test or for a

normal distribution.
:T-scores are adjusted if they failed White’s test of homeskedaticity (White 1980).
dAcross Years row reports the mean coefficient and standard deviation of the coefficient of the cross-

sectional regressions. Zl equals 1/ ‘l—N NZ” / V kj /(kj 2) wheret is the t-statistic for year 1, k is the
degrees of freedom for year j, and N rs the number of years. 22, which corrects for cross-sectional and

serial correlation, equals t/[Stddev(t)m]. (See Barth 1994.)

81

Table 4

Earnings Modela

NIBRDj.t = (11+ [31 NDAj.t-l + B2R&Dj.I-l + 133155568161 + 51.1

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Coefficient
(Standard Error)
[adjusted T or 2]“d
Panel A: Traditional Firms
NDALH R&Dj’H ASSCISJ'VH Adj. R: N
Pooled with 0.0075 0.2279 0.8499
year (0.0019) (0.0501) (0.0634) 0.98 169
intercepts [3.95]3 [4.55]3 [13.41]3
0.0122 -0.0825 1.1458
1986 (0.0033) (0.0543) (0.0748) 0.99 12
[3.69]3 [.152]1 [15.31]3
0.0049 0.2216 0.7818
1987 (0.0073) (0.1288) (0.1889) 0.97 14
[0.67] [1 .72]‘ [4.14]3
0.0053 0.2866 0.8058
1988 (0.0049) (0.0925) (0.1329) 0.99 14
[1.09] [3.10]3 [6.06]3
0.0039 0.0873 1.0284
1989 (0.0047) (0.1372) (0.1839) 0.97 13
[0.83] [0.64] [5.59]3
0.0084 0.0172 1.2237
1990 (0.0123) (0.1661) (0.2395) 0.98 14
[0.68] [0.10] [5.11]3
0.0127 0.1948 0.7730
1991 (0.0235) (0.1002) (0.1108) 0.98 14
[0.54] [1.94]2 [6.98]3
0.0180 0.2364 0.7460
1992 (0.0177) (0.1059) (0.1062) 0.98 18
[1.02] [2.23]2 [7.02]3
0.0031 0.1858 1.0885
1993 (0.0220) (0.1347) (0.1686) 0.98 18
[0.14] [1 .38]' [6.46]3
0.0115 0.1191 1.4173
1994 (0.0183) (0.1775) (0.1860) 0.97 19
[0.63] [0.67] [7.62]3
0.0087 0.4774 0.6232
1995 (0.0126) (0.1290) (0.1309) 0.98 19
[0.69] [3.70]3 [4.76]3
0.0102 0.5176 0.5386
1996 (0.0213) (0.1812) (0.1805) 0.96 19
[0.48] [2.86]3 [2.98]3
Across Years 0.0090 0.1808 0.9247
(0.0045) (0.2061) (0.2735)
21 [2.92]-‘ [4.32]3 [20.20]3
22 [3.19]3 [2.67]3 [6.44]3

 

 

 

 

 

 

82

 

Where:

Table 4 (continued)

 

 

 

 

 

 

 

 

Panel B: Generic Firms
NDALH R&Dj,r-l ASSCtSj‘H 1Adj.l{Z N
Pooled with 0.0128 0.2356 0.7423
year (0.0091) (0.1037) (0.0965) 0.86 78
intercepts [1 .41]' [2.27]2 [7.69]3
-0.0637 0.9955 0.3434
1993 (0.1173) (0.5511) (0.4278) 0.76 10
[0.54] [1.81]2 [0.80]
0.0466 0.1321 1.0028
1994 (0.0309) (0.2276) (0.2280) 0.87 12
[1 .51]l [0.58] [4.40]3
0.0302 0.2339 0.7880
1995 (0.0216) (0.1444) (0.1560) 0.93 12
[1.40]' [1.62]‘ [5.05]3
Across Years 0.0044 0.3658 0.7114
(0.0595) (0.5753) (0.3363)
21 [1.25] [1.48]' [5.40]3
22 [0.96] [1.01] [2.11]2

 

 

 

 

 

 

 

NIBRDJ;t net income from continuing operations before depreciation and amortization and
before research and development expenditures for firm j during year t;

at; year specific indicator intercepts (untabulated);

NDA)“: the number of New Drug Applications approved by the Food and Drug
Administration for firm j during fiscal year t-l;

R&DLH: research and development expenditures incurred by firm j during year t-l;

Assetsj,._.: total assets of a firm at fiscal year end t-l.

'All financial variabes, Sales, R&D, and Assets, are in millions of dollars. Further, they are included in the
model in logarithmic form.

”I, 2, and 3 indicate significance at the .10, .05, and .01 levels respectively for a one-tailed t-test.

c’T-scores are adjusted if they failed White’s test of homeskedaticity (White 1980).

dAcross Years row reports the mean coefficient and standard deviation of the coefficient of the cross-

N
sectional regressions. Zl equals 1/ WE}! / V ,9 / (k! T 2) where t, is the t-statistic for year j, k, is the
degrees of freedom for year j, and N is the number of years. 22, which corrects for cross-sectional and

erial correlation, e alsi stddev(t) J. See Barth 1994.
S ‘1“ /[ m ( l

83

Table 5

Price Model

Pricejr = (11+ BINDAj.1 7' B2R&DPS]'.1 + B3EPSj.I + BJBVPSJ’J + 81.1

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Coefficient
(Standard Error)
[adjusted T or Z]"b'c
Panel A: Traditional Firms
191314,, R&DPS“ 13st, BVPS,t Adj. RI N
Pooled with 0.3352 3.2779 6.5073 0.1515
year (0.1064) (2.1572) (1.0646) (0.2592) 0.62 212
intercepts [3.15]3 [—1.52]‘ [6.11]3 [0.58]
1.9061 -2.0825 -3.9785 3.4735
1985 (0.8728) (9.4497) (4.0675) (2.4193) 0.54 13
[2.18]2 [0.22] [0.98] [1 .44]'
0.5759 24.9604 4.1799 0.5873
1986 (0.5300) (12.5183) (4.8596) (1.9332) 0.34 15
[1.09] [1.99]2 [0.86] [0.30]
0.3957 11.8573 0.7597 1.4591
1987 (0.2891) (7.6962) (3.2556) (0.7880) 0.77 15
[1 .37]' [1.54]1 [0.23] [1.85]2
0.0332 -18.9163 16.4024 -2.6210
1988 (0.3743) (14.7830) (7.0994) (1.7793) 0.50 16
[0.09] {-1.28] [2.31]2 [.147]1
0.9731 13.7395 -7.9598 3.1985
1989 (1.1184) (17.5843) (6.4806) (1.6167) 0.15 16
[0.87] [0.78] [-123] [1.98]2
0.4424 10.4792 4.6124 2.5707
1990 (2.1504) (13.3067) (3.9380) (2.0539) 0.01 17
[0.21] [0.79] [1.17] [-125]
1.6836 1.2541 3.8668 0.1808
1991 (0.7137) (5.3972) (2.6870) (1.3590) 0.57 20
[2.36]2 [0.23] [1.44]1 [0.13]
0.2690 -5.4465 8.0286 0.1057
1992 (0.4819) (3.0969) (1.4718) (0.7103) 0.78 21
[0.56] [.1.76]2 [5.45]3 [0.15]
0.2155 -5.0196 7.5232 0.0580
1993 (0.2658) (2.5912) (1.4142) (0.4320) 0.80 21
[0.81] [.194]2 [5.32]3 [0.13]
0.0527 -l0.0805 10.2361 0.2520
1994 (0.2202) (2.5656) (1.0950) (0.3924) 0.92 21
[0.24] [.393]3 [9.35]3 [0.64]
0.8766 -6. 1764 1.6934 1.4299
1995 (0.9348) (9.3545) (3.7536) (1.3892) 0.04 22
[0.94] [0.66] [0.45] [1.03]

 

 

 

 

 

 

 

84

 

Table 5 (continued)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

NDA“ R&DPS“ EPSL, BVPS], Adj. R” N
0.4809 21.9735 14.8008 1.6185
1996 (0.6494) (5.1654) (2.7228) (0.7898) 0.75 22
[0.74] [.425]3 [5.44]3 [2.05]3
Across Years 0.6228 -0.6l71 4.1905 0.5976
(0.6443) (13.8131) (7.6250) (1.8766)
21 [2.66]3 [2.41]2 [7.60]3 [1.90]2
22 [3.07]3 [-121] [2.28]2 [1.66]2
Panel B: Generic Firms
Pooled with 0.4733 0.0107 4.1340 0.1185
year (0.1110) (1.6864) (1.3068) (0.3950) 0.39 120
intercepts [4.26]3 [0.01] [3.16]3 [0.30]
1.3269 .1.5113 16.6588 .1.7025
1992 (1.1047) (10.3764) (12.0474) (2.7394) 0.56 13
[1.20] [0.15] [1.38]' [0.62]
0.2511 4.8993 9.1440 -1.7680
1993 (0.3815) (7.2803) (6.8099) (2.2438) 0.07 13
[0.66] [0.67] [1.34] [0.79]
0.5549 .3.3725 7.9339 0.0438
1994 (0.2527) (2.7425) (2.5157) (0.8449) 0.79 15
[2.20]2 [-123] [3.15]3 [0.05]
1.4072 0.0172 0.7089 1.4991
1995 (0.6741) (8.2967) (4.9908) (1.4294) 0.51 15
[2.09]2 [0.00] [0.14] [1.05]
0.8680 0.7826 5.2868 0.7228
1996 (0.4209) (4.2754) (4.0313) (1.2417) 0.40 14
[2.06]2 [0.18] [1.31] [0.58]
Across Years 0.8816 0.1562 7.6629 -0.5301
(0.4947) (3.0884) (6.3033) (1.3593)
21 [3.40]3 [0.22] [2.92]3 [0.36]
22 [4.83]3 [0.30] [2.41]3 [0.47]

 

 

 

 

 

 

 

85

 

Where:
Price“:
01,:
NDAJ'J:
R&DPS“:

BPS“:

BVPS)“:

Table 5 (continued)

stock price for firm j at the end of the first quarter following the fiscal year end;
year specific indicator intercepts (untabulated);

the number of New Drug Applications approved by the Food and Drug
Administration for firm j during fiscal year t;

research and development expenditures incurred by firm j during year t divided
by the number of shares outstanding;

net income from continuing operations before depreciation and amortization and
before research and development expenditures for firm j during year t divided by
the number of shares outstanding;

total assets minus total liabilities for firm j at the fiscal year end t divided by the
number of shares outstanding.

8I, 2, and 3 indicate significance at the .10, .05, and .01 levels respectively for a one-tailed t-test.
l’T-scores are adjusted if they failed White’s test of homeskedaticity (White 1980).
cAcross Years row reports the mean coefficient and standard deviation of the coefficient of the cross-

N
sectional regressions. Zl equals 1/ Wilt} / ‘1’ lg /(kj _ 2) where t,- is the t-statistic for year j, k,- is the

degrees of freedom for year j, and N is the number of years. ZZ, which corrects for cross-sectional and

serial correlation, equals t/(Stddefii2/N

I). (See Barth 1994.)

86

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

E8 5?: 88-: :48 5.: 5?:
2 8.? 58.8 $8.8 $1.8 $88 8288 A888 82
:88 $8.? 88.? 28.8 2.8.8 R8,?
.80: :28 :81 :2: .38.: :88:
2 :8 85.8 288.8 282.8 842.8 58.8 288.8 82
2.28 88.8 28.? 82.? 88.? 8o?
58 5.8 $3-: 80?: .88 so:
2 2.? 2288 $8.8 8.2.8 2828 628.8 288.8 82
88.8 88.8 88.? 38.? 25.8 88.8
5.: .82-: $2-: .5: 5.: :n:
2 2.? 2228 :38 8298 6:28 588 858.8 88
88.8 22.? 82.? some 888 88.8
.E: :28 5.1 5?: 5.8 88?:
2 8.8 88:8 388.8 88.8 52.8 228.8 88.8 82
88.8 88.8 $8.? 28.? 38.8 e88?
22?: $3-: .88-: :28 .88 5.8
2 on? E88 888 2828 $88 8288 888.8 52
$8.? $8.? 88.? .8 3 85.8 88.8
8?: 5?: 38?: 72?: 2.8 .28-:
z 8.? 288.8 698.8 82.8 288.8 88.8 35.8 82
R8? 88.? 2 _ 3. 28.? 28.8 $8.?
”.38 83-: 45.3 :28 808 $8-: 3885
m8 :8 628.8 5.88 £888 888 888 A888 .8»
28.8 38.? 82.? 88.8 28.8 28.? .83 Boom
z .013. 4.42% :22 $3032 4.2583. 4.6224 ....<az
math 35.3595. 3‘ Wash
aim co 9 8.3.68
coca 88:88
Ens—beau

 

 

 

.26 + .2893 + 2258 +...maoex<£ + 488328 + 25248 + 232.8 + .o 1:833.

682 £553:

9 039—.

87

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

_:m.: .55: 28?: E: 28?: :88
2 5.5 8808 $58 $888 $298 $258 58.8 82
28.5 2.8.? 55.? 2.2.5 85.? 28.5
.55-: 55.8 .88 :81 :28 5.8
2 5.? 542.8 2288 55:: 52.8 988.8 :88 82
28.? 58.5 2.8.5 :2? 25.5 55.5
N52: :55: .88 :52 52-: :5:
2 25 E88 528.8 555.8 8%.: 2558 3.2.8 82
:55 88.? 42.3. 3.3- 9.2.? 588
.5: $2-: N55.1 :88 52-: .52-: 88.85
2: 2.5 55.8 58.8 5.58 222.8 2288 E88 .82
548 55.? 52.? 85.5 88.? 2 5.? 8.... oozed
2 R 8.... 4.4254 4.25 4:882 4.83.2 ...<QZ< 352
mgr—mn— utoflumv “Q 3:!—
.58 53-: .53-: 5?: .22.: 5.8 Q
.58 52-: ”5.1 8?: .5: 54.8 _N
5.28 28.8 5E8 858 255.8 2858
88.5 88.? ER? 58.5 38.5 885 88> cocoa.
28?: NEC: H28 .53-: .85-: :21
a 85 $588 5588 2:28 328 8.5.8 5.5.8 82
88.? 825 85.5 88.? 585.? 88.?
521 50?. :28 5.8 :58 .88
5 8.? 53.8 58.8 53.08 22.8 55.8 5.58 82
28.? 88.? 28.5 88.5 28.5 28.5
NS: :88 :58 :21 :28 :2:
8 25 58.8 325.8 88.8 525.8 358.8 585.8 52
22.5 85.5 855 58.? 885 555
45.8 N:92 :2: NE: 22.8 :21
a 2.5 6.8.8 $8.8 38.8 $25.8 535.8 525.8 52
82.5 :8? ES- 22.5 885 R8?
z a :3. 22:2 .25 4.82632 4.2552 .4524. ....<oz

 

 

 

 

 

 

 

 

 

 

€85.88 c 2...?

88

Adam. 5.5m. 8m. . £38358 . fiasco dorm—2.50 .888 28 35.88.80...
“x
5.. 38:8 :03? .NN .859. ..o .6353: 05 m. 2 :5 <8?» 3.. 6032.. Mo 38mg 05 m. .5. <82. 5.. 0.8.58».-. 05 m. x. 20:? I s \ LN.
.5 83% . 2 2?.

£533 .N 8.5.833. 3.858-880 2.. .3 80.0508 05 we 8:530: .5898:- ES 55.0508 :38 05 $888 32 83> 880$

.83. 85>». 3.0333889. no 88 ESE? 3.8.. .35 w. 383.85 25 888....
.58-. 3.5-0.8 m .8 8.3.82.8. 295. .c. .55 .mo. 6.. ofi .5 8:85:83 286:. m .55 .N .r

.5 3?. o. 7. Bo» Bo... mam a. omen—.0 2.. ”2mmm4
852583 8.82.... 2.. .3 3.5.3.5 . .59. mats:
82365on0 5:25.238 98 £25.82 282. 98 3.88825 98 5.8.52ro 282. 8.8.8qu wfisfiaoo Soc 088:. .0: “imam
u 89. o. .8 89. Boa mmafim E owzmno 05 Hammad”?
85233.5 88am .o 885.: 05 .3 .8636 5 So» war—.6 358:. 8.8659... 5588238 98 £883. “immafim
u.-. .88 5 89. .58... 59,32. ..- an... 5.. <9. 05 .3 3.63% aconao..&< man. 26 Z we 538:: o... E uwcuno 2.. 2.2524
3 So» .58... wan-:6 ..- an... 3.. 8.852352 men .55 too". 95 3 .5268...“ 38.58.89... wan. >62 .3 898:: 05 n...<QZ
.€o.m.:p§§. 35.0828 8.86:. 2.28% 89. “so
..+. 89. 8.. Stan: .8... 2: mo .55 2.. 8 . SP. 8.. 8.83.5 .8... 2.. -.o .55 2.. 80.... .8 SE 8.. EBB 58.88 H...anm
88:3

 

 

 

 

 

 

 

 

 

 

 

 

M.88 .85-. .85. .55-. .85-. .88 NN
n.85. .85-. .85. .85-. .85-. .85. .N
.888 .888 .288 .M: 5. 8 :8 _ 8 .888
585 8:5- 52... 88.? 88.? 885 38:85..
.35.... .88 .8.-. .85. .85-. .85-.

2 85 .888 .828 .888 .888 .888 .5858 82
3.85 885 $85- 885 88.? 88.?

z w. .3. .88 :8”. 8.555 .893. ....<az< ...<az

 

 

€35.58. 5 25¢

89

Table 7

Sales Model with Patents (Traditional Firms Only)

Sales“ = a! + BINDAj‘H + BzPatentsJ-J-x + B3R&Dj.t-x + B4ASSCtSj'H + 8]".

 

Coefficient
(Standard Error)
[adjusted T]“’b’c

 

NDA).1-1

0.0110
(0.0027)
[4.07]3

0.0234
(0.0080)
[2.93]3

 

Patents”,

0.0010
(0.0006)
[1.67]2

-0.0023
(0.0025)
{-0.92}

 

Patents”;

-00007
(0.0034)
[-021]

 

Patentsjw

-0.0022
(0.0035)
{-0.63]

 

PatentsJ-H

0.0041
(0.0042)
[0.97]

 

Patents)“

0.0032
(0.0036)
[0.89]

 

R&Dj.t-I

-0.1 190
(0.0498)
{-2.39}3

-O.7673
(0.1971)
{-3.89}3

 

R&D)“

0.6516
(0.2466)
[2.64]3

 

R&Dj.t-3

-0.0989
(0.2237)
{-0.44}

 

R&D)“

0.1578
(0.2270)
[0.70]

 

R&D)“

-0.0157
(0.1579)
[-010]

 

Assets”.

1.1660
(0.0592)
[19.67]3

1.0832
(0.0823)
[13.16]3

 

Adj. R‘

0.96

0.96

 

 

N

 

185

 

103

 

9O

 

Where:
Sales)-J
(1.:
NDAJJ-“
Patentsm:

R&Dj_.-.:

Assets“- .2

Table 7 (continued)

revenues for firm j during year t, excludes interest other non-operating revenues;
year specific indicator intercepts (untabulated);

the annual number of New Drug Application approved for each firm by the
Food and Drug Administration for year t- l;

the number of patents obtained by firm j during year(s) t-x, x ranges from one to
five years;

research and development expenditures incurred during year(s) t-x, x ranges
from one to five years;

total assets of a firm at the fiscal year ended t-l.

aAll financial variabes, Sales, R&D, and Assets, are in millions of dollars. Further, they are included in the
model in logarithmic form.

l’1, 2, and 3 indicate significance at the .10, .05, and .01 levels respectively for a one-tailed t-test.

c’1"-scores are adjusted if they failed White’s test of homeskedaticity (White 1980).

91

Table 8

Earnings Model with Patents (Traditional Firms Only)

NIBRDN = (I! + BlNDAjJJ + [3213316113534 B3R&Dj.t-x + B4ASSCTSLH + 8}"!

 

Coefficient
(Standard Error)
[adjusted T]“‘"'°

 

NDA).-.

0.0077
(0.0018)
[4.28]3

0.0141
(0.0050)
[2.82]3

 

Patentst

0.0008
(0.0004)
[2.00]2

0.0006
(0.0016)
[0.38]

 

PatentsJ-H

00009
(0.0021)
{-0.431

 

Patents)“

0.0001
(0.0022)
[0.04]

 

Patentst

0.0008
(0.0027)
[0.30]

 

Patents“;

-0.0001
(0.0023)
{-0.04}

 

R&D)“

0.2218
(0.0495)
[4.48]3

0.1063
(0.1519)
[0.70]

 

R&Dj.t-2

-O.484l
(0.2120
{-2.28}

 

R&Dj.t-3

0.5900
(0.1548)
[3.81]3

 

R&Dj’t4

0.3033
(0.1514)
[2.00]2

 

R&D).-.

-O.l842
(0.1052)
{-1.75}2

 

Assets”.

0.8409
(0.0632)
[13.31]3

0.6764
(0.0569)
[1 1.89]3

 

Adj. RI

0.98

0.98

 

N

 

 

169

 

101

 

92

 

Table 8 (continued)

Where:

NIBRDJJ net income from continuing operations before depreciation and amortization and
before research and development expenditures;

(1‘: year specific indicator intercepts (untabulated);

NDA)“: the annual number of New Drug Application approved for each firm by the
Food and Drug Administration for year t-l;

Patentsj..-,(: the number of patents obtained by firm j during year(s) t-x, x ranges from one to
five years;

R&Dj“: research and development expenditures incurred during year(s) t-x, x ranges
from one to five years;

Assetsj,.-.: total assets of a firm at the fiscal year ended t-l.

aAll financial variabes, Sales, R&D, and Assets, are in millions of dollars. Further, they are included in the
model in logarithmic form.

b1, 2, and 3 indicate significance at the .10, .05, and .01 levels respectively for a one-tailed t-test.

c’1"-scores are adjusted if they failed White’s test of homeskedaticity (White 1980).

93

Table 9

Price Model with Patents (Traditional Firms Only)

Price“ = at + Bl NDA“ + BZPatentst-x + B3R&DPSj.t-x + B4EPSj.1-l + BSBVPSj.t-l + 8111

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Coefficient
(Standard Error)
[adjusted T]“b
NDA“ 0.2783 0.6337
(0.0986) (0.2535)
[2.82]3 [2.50]3
Patents“ 0.0630 —0.0987
(0.0256) (0.0729)
[2.46] {-1.35]
PatentsJ-‘H -0. 1004
(0.1 107)
{-0.91]
Patents)“ -0.1 1 11
(0.1162)
{-0.96]
Patents-(-3 0.0788
(0.1 184)
[0.67]
Patents)“ 0.3039
(0.12113)
[2.51]
R&DPSjJ -3.0656 -10.5340
(1.96)l (3 .07903)
{-1.56] {-3.42]
R&DPS].-. -1.2601
(3.1755)
{-0.40]
R&DPS”; 7.6731
(3.46313)
[2.22]
R&DPS)“ 1.9471
(3.5683)
[0.55]
R&DPS)“ 6.3086
(2.8271)
[2.23]3
EPSJ;l 4.0885 1.1591
(1.10023) (1.1571)
[3.72] [1.00]
BVPS” 0.3309 1.2681
(0.2597) (0.39263)
1.27 ' 3.23
Adj. Rz [0.64] [0.53]
N 212 129

 

94

 

Where:
Price“:
on:
NDA“:

Patents)“:

R&DPS-J-,:

BPS)“:

BVPS“:

Table 9 (continued)

stock price at the end of the first quarter following the fiscal year end;

year specific indicator intercepts (untabulated);

the annual number of New Drug Application approved for each firm by the
Food and Drug Administration for year t;

the number of patents obtained by firm j during year(s) t-x, x ranges from zero
to four years;

research and development expenditures incurred during year(s) t-x, x ranges
from zero to four years, divided by the number of shares outstanding;

net income from continuing operations before depreciation and amortization and
before research and development expenditures during year t divided by the
number of shares outstanding;

total assets minus total liabilities at the fiscal year end t divided by the number
of shares outstanding.

‘1, 2, and 3 indicate significance at the .10, .05, and .01 levels respectively.
bT-scores are adjusted if they failed White’s test of homeskedaticity (White 1980).

95

Table 10

Successfiil Price Model

Price]. = a. + BISuccessj + [32R&DPSj_, + B3Successj‘R&DPSJ-J_. + B41:‘.PSJ-_l

+ 13st31.1 + 811

 

Coefficient
(Standard Error)
[Adjusted T or Z]"b‘°

 

Panel A: Traditional Firms

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Successj R&DPS“ Successj" BPS]. BVPS];l Adj. R‘ N
R&DPS]...

Pooledwith 11.2955 -1.4585 6.2564 2.9752 0.3615

year (3.4029) (1.4661) (1.7972) (0.8116) (0.2647) 0.71 212

intercepts [3.32]3 {-0.99] [3.48]3 [3.67]3 [1.37]l
32.8540 34.1536 -33.0876 2.0491 0.0767

1985 (21.3188) (23.5194) (22.5664) (3.8030) (2.7200) 0.41 13
[1.541' [1.451' {-1.47]' [0.54] [0.03]
25.8607 35.8060 46.0443 -3.3709 0.1281

1986 (31.9377) (47.8843) (48.2880) (5.4270) (2.0888) 0.23 15
[0.81] [0.75] {-0.33] {-0.62] [0.06]
18.1761 30.3771 -24.0469 1.3384 1.0104

1987 (13.8960) (17.3307) (17.3776) (3.2425) (0.8294) 0.76 15
[1.31] [1.75]' [-1.38]' [0.41] [1.22]
40.4523 0.6786 -27.1332 17.4236 .3.3794

1988 (18.4346) (15.6044) (16.7564) (6.2004) (1.5134) 0.63 16
[2.19]2 [0.04] [-1.62]' [2.81]3 {-2.23]2
30.0187 11.2218 -6.2697 -6.0489 2.7849

1989 (22.0731) (19.2138) (20.8108) (5.7159) (1.6328) 0.17 16
[1.36]' [0.58] {—0.301 {-1.06] [1.71]2
16.9116 10.9238 -9.1787 5.8084 -2.2167

1990 (26.6284) (14.8157) (19.0969) (6.2865) (2.3145) —0.05 17
[0.64] [0.74] {-0.48] [0.92] [-0.96]
11.9319 -5.6313 14.8592 -0.4767 1.9738

1991 (10.6385) (4.5101) (6.1767) (2.4162) (1.1663) 0.76 20
[1.12] [-125] [2.41]2 {—0.20] [1.69]2
7.0286 -3.8259 3.5433 5.2540 0.3833

1992 (6.3974) (2.8763) (4.2670) (2.1115) (0.6827) 0.82 21
[1.10] {—1.331' [0.83] [2.49]3 [0.56]
3.1609 —2.1893 3.9526 4.2536 0.2042

1993 (6.5987) (2.6717) (2.7982) (1.4657) (0.4611) 0.85 21
[0.48] {-0.82] [1.41]' [2.90]’ [0.44]
-2.8063 —5.9245 10.4448 4.6895 0.9994

1994 (4.7911) (2.2232) (2.5015) (1.3557) (0.3385) 0.97 21
[-059] [.266]3 [4.18]3 [3.46]3 [2.95]3
41.3031 5.3208 -9.3064 -0.6947 0.0723

1995 (21.1936) (11.4374) (10.7519) (6.0477) (1.5184) 0.07 22
[1.95]2 [0.46] {-0.87] {—0.1 1] {-0.05]
11.1120 47.5444 6.1405 11.5476 1.5419

1996 (10.1224) (3.9274) (5.9947) (2.0930) (0.6502) 0.86 22
[1.10] {—4.47]3 [1.02] [5.52]3 [2.37]2

 

 

 

 

 

 

 

 

96

 

Table 10 (continued)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Successj R&DPSJ-J Successj“ EPSJ;1 BVPS“ Adj. R1 N
R&DPS,“
Across Years 19.6670 7.7805 -7. 1772 3.4811 0.2862
(14.4588) (17.3544) (15.5343) (6.3789) (1.6953)
21 [3.53]3 [4.341' [0.97] [4.67]3 [2.14]2
22 [497]" [-073] [0.54] [2.37]3 [1.49]'
Panel B: Generic Firms

Pooled with 8.0788 7.3052 41.7805 4.9133 .0209

year (1.8767) (4.9662) (5.3833) (1.3392) (0.4472) 0.41 118
Intercepts [4.30]3 [1.47]‘ [.219]2 [3.67]3 [-049]
-8.8603 .53.4245 65.2241 32.2099 -5.1361

1992 (11.0108) (68.1486) (70.6562) (10.7711) (2.4496) 0.47 13
{-0.80] {-0.78] [0.92] [2.99]3 [.210]2
5.7964 12.3656 -8.3423 9.7774 -2.1018

1993 (7.9234) (41.4010) (40.2763) (7.5280) (2.3903) -003 13
[0.73] [0.30] [-021] [1.30] {-0.88]
8.0935 12.3984 49.3259 8.3275 0.2928

1994 (4.3905) (16.6238) (17.0571) (2.8165) (1.0183) 0.75 15
[1.84]2 [0.75] [4.13] [2.96]3 [0.29]
7.1544 24.4275 -35.8778 4.7375 3.9690

1995 (6.1750) (15.9983) (16.1250) (5.2743) (1.6959) 0.49 15
[1.16] [1.53]2 [.222]2 [-033] [2.34]2
6.7785 5.8685 43.2178 6.1720 0.3098

1996 (6.4760) (8.2568) (10.8082) (4.7644) (1.4874) 0.18 14
[1.05] [0.71] [4.22] [1.30] [0.21]
Across Years 3.7925 0.3271 -2.3079 10.9499 -0.5333
(7.1209) (30.7872) (39.1553) (12.6889) (3.3656)
21 [1.65]1 [1.04] [4.61]' [3.39]3 [-004]
22 [1.621' [1.18] [4.30]' [2.37]3 [-003]

 

97

 

 

Where:
Price“:
(1]:
Success]:

R&DPS”:
Success“

R&DPS)“:
raps”:

BVPSj,tI

Table 10 (continued)

stock price at the end of the first quarter following the fiscal year end;

year specific indicator intercepts (untabulated);

a dummy variable that is set to 1 if a firm’s mean NDA’s (across the sample
period) is greater than the median of all firms of the same type;

research and development expenditures incurred during year t divided by the
number of shares outstanding;

the success dummy variable multiplied by R&DPS during year t-l;

net income from continuing operations before depreciation and amortization and
before research and development expenditures during year t divided by the
number of shares outstanding;

total assets minus total liabilities at the fiscal year end t divided by the number
of shares outstanding.

“I, 2, and 3 indicate significance at the .10, .05, and .01 levels respectively for a one-tailed t-test.
bT-scores are adjusted if they failed White’s test of homeskedaticity (White 1980).
cAcross Years row reports the mean coefficient and standard deviation of the coefficient of the cross-

N
sectional regressions. Z1 equals 1/ mg!” / V kj/ (k1 - 2) where I) is the t-statistic for year j, k, is the

degrees of freedom for year j, and N is the number of years. Z2, which corrects for cross-sectional and

serial correlation, equals /(S‘ddev(’)//

1)‘ (See Barth 1994.)

98

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

~58 88.8 .88.: 88.: N52 .8: 5.8
8 28 888.8 38.8 248.8 38.8 58.8 588 88.8.8 88
88.8 88.8 883 238- 8:8- 888 88.8
:31 E8 ”88.1 .5: .5: :61 $3-.
2 8.8 8828 $888 58.: 88>: E88 2 :8 38:8 38
:88- 288 38 m- 483 85.8 88.8- 88.8-
.8.3 821 88.8 88.1 35.2 382.8 88.8
2 «no 8888 888.8 58.: 688.: A888 8.8.: €8.88 88
38.8- 288- 88% 88 e- 28 m- 888 88.8
5.8 821 88.8 :2: 88.3 88.8 3.8
2 $8- $83.8 388.8 3358 888 A828 : 5&8 $38 $8
888 88 o- 858 8% o- 8: o- 88.8 m: 3
898 82-. ”:88 “5.1 NEE L88 NE:
2 88 88.8 58.8 3888 $88 $808 588 A838 88
3.88 :88- 882 3:8. 288- 88.8 88.8
:08 83: 88.8 58-8 5.8 8:: .82:
2 58- 65.8 88—8 58.8 3808 688.8 888.8 888.8 88
88.8 88 o- $5.8 88.7 882 :88- 88 o-
5.3 88.2 a : 32 L88 L8,: ”:28 88-3
3 8.8 A888 288.8 88.68 :38 $84.8 $58 $888 88
:88- 88 .o- 88 w 88: 88 m- 888 88.0-
888 .88 5.8 ”82.2 5:: 5.8 8.3 8855
8. :8 58.8 838.8 88_.8 82.8 58.8 888.8 288.8 a:
888 :88 :88 888- 888- 288 88.8- .83 8.88
_._mao§< 2892..
z .8 .3 flan-Z 88m .8825 _._.8a§2 .8865 36-362 .8825

 

 

mELE REE—:2... 3.. .23.—

 

 

...J + 4.8318 + 4.2-”Han + 48842418883 twang/48 + 4.2383858 + 4.298an + 88825.8 + .8 14.6368

.322 mEBoM Emmooosm

: ozah

99

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

888 .881 : .1 888 E8 5.88 58.8

2 88.8 8888 88.8 8888.8 688.8 888.: 88.: 288.8 88
88.8 888.8- 888- 888.N :82 88 8. 28.8-
8888 3.8.: .88.: ENE 888.8 88.8 .88

8 8.8- 8888.8 8888 888.8 888.8 58.8 388.8 888.8 88
:2 .8- 8 8.8 88.8 288.8- :8.N- 888 $88
88.: 88.1 :88 88.8-8 88.8-8 8.8 .888

8 :8- 8888.8 888.8 882.8 52.8 $88.8 $88.8 888.8 88
888.8 888.8- 838 888.8- 883.- 888 888.8
M7.8.8 88.8.8 7.8.8.8 888 88.8 888 5.8.8 388.25

8_ 8.8 3888.8 288.8 3.8.8 @888 88.8 688.8 688.8 .8;
88.8 88.8- 388- :888 88.8 8: 8.8 888.8- 883 8288

8:58 orgoc "m .955

88.: .52 8.8-8 88.8.8 22.. .5: 88.8 S
N88.: E82 88.8-8 E88 888.2 888.8 58.. _N
88.8 388.8 388.8 828.8 388.: $88.: 888
E88 :88- 888 888.8- 8888- 88.8 888.8- 88:83.34
88 NE: 88.1 .88.: 88.8 88.3 88.8-8

8 8.8 38.8 588 A888 9888.8 888.8 88.8 688.8 88
88.8 8 8.8 888 8- 8888 888.8 888 8- 88 8.8-
8888 .888. _ -8 8888 78.8.8 :88 .88.: : 2-8

a 8.8 388.8 288.8 @888 888.8 88.8 888.8 9888.8 88
88 8- 82.8- 8888 8:8- 888 58.8 88.8-
888: 8.8 888.8 .888 :88 88.8.8 888.8

8 _8.8- 388.8 588.8 88.8 9888 88:8 88:8 6888 88
8 8.8 888.8 888.8 888.8- 88888 888 8- 888.8
88.8 8.8.1 .8: E82 .88 L88 5.8-8

8 88 A888 8888 :8 8.: 38.8. : 888.8 888 E88 88
88.8 88.8- ~82 888.8- 888.8 88.8 88.8-

.8582 _. 8888.8.
2 .8 .834 4.888 88-8 r8885 883.32 888:8 4.8988 888:8

 

€388.38 : 28.:-

 

 

:88 858 668 .

T 2
A8888:

8 8133 62822808 3.8 was .8283.

-888 88 88:8 523 .NN .8898 80 838:: 05 2 2 can .338 3.8 Eovoot mo 82mg 05 818‘ <88» 8o.“ 83388.8 05 8 .8 223 An I .8? .8 > \ mum-8.8% 2

8.833 _N 8888088 3.858-880 05 .«o 856888 05 MO 82338 98588 98 856508 :88 05 $8082 38 88“; 8885

.888 8:33 3685888882 80 888 8.833 3:88 805 a 883.288 Ba 8288.-.:
.82-. 3:88-28 8 8888 E35388 296— 8. 98 .mo. .2. 05 88 3:85:88 8865 m 98 .N . r

.7. con—B 398 38c 85 .8 3E :58 85 43 8388 8 use» 9 78 88» Bob 22?. Ba 2:85 3: E omfifio 05

:8 328 So» Roma 88 388 :58 05 an .8288 $59535 8828 .80 83:3: 2: E 32.86 8 83.8

wctzc 8888:8598 85882268 85 £88882 «.5qu 28 5.5888888 23 5:88:38 283 8838825 wag—5:8 89a 0885 S:
a 89» 9 _ 18 83» Set 95832 83 3:33.: .8 81:8 88 03889» 833m

3 .89» 9 _ I 8 .898 89a mamas”— E owfifio 2:

$53888 38:8 80 83:5: 86 .3 323—8 8 .89» mar—86 @8585 8836585 85882308 98 £288.
$.88: 80 828 2:88 05 Mo 5608 85 :85 88mb 28 2852 :82: 8.8-5 8 .x _ o. 888 8 .85 29E? 8556 8
“328—8585; 89885 89865 2.28% 88»

A .85 some 88 8 88 55:8 88:18 85 we can 85 9 T8 88 333v 88¢ Bu .8 8:0 85 89c EBB 88—88

”.8882

”888

838898.84

.8826
8888.88
8832
”88325
”so
ungom

”285$

 

 

 

 

 

 

 

 

 

 

 

 

 

.5: 888.8 3.8 888-8 888-8 :8: :8: S
.88.: 88.8-8 .888 8888-8 8888-8 :88 :88 8
888.8 288.8 888.8 388.8 88.8 $88.8 388.8
888.8 88.8- 8:5 882. :8:- 885 8888. $3,822
88.8 :88 z _.8 88.8-. 8888.8 88.8 88.1

2 88 $888 588 8:88 8888 3.8.: 822.: @8888 88
88.8 888 888.8 82;- 8.888. 88.8 888.8-

4,888.82 88988..
2 am .82 8882 888 4888688 883.82 88888 .8838 886888

 

 

88888888 : as;

101

"Tlllfii'ifl’flflflifl@‘Wfl'fli’fiflifli‘s