Digitized by the Internet Archive
in
2011 with funding from
Boston Library Consortium
Member
Libraries
http://www.archive.org/details/marketsizeininnoOOacem
iDI
^b\
HB31
.M415
33
Massachusetts Institute of Technology
Department of Economics
Working Paper Series
Theory and Evidence
from the Pharmaceutical Industry
Market
Size in Innovation:
Daron Acemoglu
Joshua Linn
Working Paper 03-33
September 2003
Room
E52-251
50 Memorial Drive
Cambridge,
MA 021 42
paper can be downloaded without charge from the
Social Science Research Network Paper Collection at
This
http://ssrn.com/abstract=457440
Market Size
in Innovation:
Theory and
Evidence From the Pharmaceutical Industry*
Daron Acemoglu
Joshua Linn
MIT
MIT
September 2003.
Abstract
This paper investigates the effect of (potential) market size on entry of new drugs and
pharmaceutical innovation. Focusing on exogenous changes driven by U.S. demographic trends,
we find that a 1 percent increase in the potential market size for a drug category leads to a 4 to
6 percent increase in the number of new drugs in that category. This response comes from both
the entry of generic drugs and new non-generic drugs, and is generally robust to controlling for
a variety of non-profit factors, pre-existing trends, and changes in health care coverage.
Keywords: Economic Growth,
JEL
Entry, Innovation, Pharmaceuticals, Profit Incentives.
Classification: 031, 033, L65.
*We thank David
Autor, Ernie Berndt, Olivier Blanchard, Alfonso Gambardella, Amy Finkelstein, Stan
at the NBER Summer Institute
and Wharton for useful comments. We are especially grateful to Frank Lichtenberg for generously sharing
the Food and Drug Administration drug approval and the Computer Retrieval of Information on Scientific
Projects data with us. This research is partially funded by the National Science Foundation and the Russell
Sage Foundation.
Finkelstein, Jerry
Hausman, Paul Joskow, Frank Lichtenberg and participants
Introduction
1
This paper constructs a simple model linking innovation rates to current and future market
and provides evidence from the pharmaceutical industry to support
size,
this hypothesis.
Our
empirical work, which exploits changes in the market size for various drug categories driven by
U.S. demographic trends, finds economically significant and relatively robust effects of market
on entry of new drugs.
size
Although many
historical accounts of
important innovations focus on the autonomous
progress of science and on major breakthroughs that take place as scientists build on each
1
other's work, economists typically emphasize profit incentives
size of the target
mar-
For example, in his seminal study, Invention and Economic Growth, Schmookler argued
ket.
that: "...invention
sued
is
largely
for gain" (1966, p.
of his chapters " The
The
an economic
206).
activity which, like other
To emphasize
amount of invention
role of profit incentives
is
the role of market
and market
change
new
1965,
make
(e.g.,
important both for the
profit incentives the central driving
Aghion and Howitt, 1992, Grossman
induced innovation and directed technical
for the
which investigate the influence of
literatures,
activities, is pur-
Schmookler entitled two
size,
size in innovation is also
force of the pace of aggregate technological progress
and Helpman, 1991, Romer, 1990), and
economic
governed by the extent of the market."
recent endogenous technological change models, which
of
and the
profit incentives
on the types and biases
technologies (see, for example, Kennedy, 1964, Drandkis and Phelps, 1965, Samuelson,
Hayami and Ruttan,
1970,
and Acemoglu, 1998, 2002, and 2003).
A
recent series of
papers by Kremer, for example (2002), also build on the notion that pharmaceutical research
is
driven by market size and argue that there
for third-world diseases
In this paper,
of
new drugs and
we
generally insufficient research to develop cures
such as malaria.
investigate the effects of market size for different types of drugs
innovation.
A
major
size
on innovation
Our
strategy to overcome this problem
is
is
difficulty in
any investigation of the impact of market
—better products
the endogeneity of market size
is
on entry
will
have bigger markets.
to exploit variations in market size driven
by demo-
graphic changes (or past demographic trends), which should be exogenous to other, for example
scientific,
determinants of innovation and entry of
new
drugs.
2
We
create the potential market
'See, for example, Ceruzzi (2000), Rosenberg (1974) and Scherer (1984). Ceruzzi emphasizes the importance
of a number of notable scientific discoveries and the role played by certain talented individuals in the development
of modern computing.
He points out that important developments took place despite the belief of many
important figures in the development of the computer that there would not be a market greater than a handful
of personal computers in the United States (2000, p. 13).
2
For many drugs non-U. S. markets may also be relevant. Nevertheless, the U.S. market is disproportionately
important, constituting about 40 percent of the world market (IMS, 2000). Below we also look at the impact
of West European and Japanese demographic changes on the entry rates of new drugs.
size for various
drug categories according to the age distribution of their users at a given point
and then trace changes
in time,
market
in potential
by changes
size driven
in
demographics
3
(holding the age profile of consumption of various drug categories constant over time).
We
measure entry and innovation using the Food and Drug Administration's (FDA) approval of
new
drugs.
Our
4
results
introduction of
show that there
new drugs
to
is
an economically and
market
size.
statistically significant response of the
For example, a
percent increase in the size of the
1
potential market for a drug leads to a 4 to 6 percent increase in the
check the robustness of our results by controlling for lagged
number
FDA approvals,
of
new
generics
and non-generics, and we
of generics
We
is
larger
and changes
whether
it
is
new
market
size
We
to past market size.
have the strongest
effect
predictions of the theoretical model
we
On
drugs.
though the response
may
the one hand, because changes in
respond to anticipated future market
the other hand, because the development process of
may respond
size,
the current market size or past or future market sizes
demographics are known in advance, drug entry
On
market
precisely estimated.
that have the largest effect on entry of
size.
drugs that enter the market comprise both
find that both respond to
and somewhat more
also investigate
New
drug expenditures.
on entry
we use
new
rates of
new drugs can be
market
find that current
size
is
consistent with the
to motivate our empirical investigation.
market
size,
and
but statistically insignificant relationship between future anticipated market
reflect the difficulty of
statistical association
innovation, on the other,
The
classic
Finally,
find a positive,
size
and patents,
matching patents to drug categories.
There are a number of other studies related to our work.
ments a
long, entry
and 5-10 year leads of
drugs, which
also look at the response of pharmaceutical patents to
which might
We
pre-existing trends,
differences in the non-economic incentives to innovate, advances in biotechnology
in insurance coverage of
drugs.
between investments and
sales,
and argues that the causality ran
First,
Schmookler (1966) docu-
on the one hand, and patents and
largely
from the former to the
latter.
study by Griliches (1957) on the spread of hybrid seed corn in U.S. agriculture also
provides evidence consistent with the view that technological change and technology adoption
are closely linked to profitability and market size. Pakes and
this issue using
factor
a more structural approach, linking
demands and
to growth of output.
In
R&D
more recent
Schankerman (1984)
investigate
intensity at the industry level to
research, Scott
Morton (1999) and
3
Loosely speaking, "market size" corresponds to the number of users times their marginal willingness to pay.
Therefore, market size can increase both because the number of users increases and because of their marginal
willingness to pay changes. We focus on changes driven by demographics to isolate exogenous changes in market
size.
'These data were previously used by Lichtenberg and Virahbak (2002), who obtained them under the Freedom
We thank Frank Lichtenberg for sharing these data with us.
of Information Act.
Reiffen
and Ward (2002) study the decision of firms to introduce a new generic drug and
a positive relationship between entry into a new market and expected revenues
None
market.
size,
find
in the target
of these studies exploit a potentially exogenous source of variation in market
however.
Second, some recent research has investigated the response of innovation to changes in
energy prices. Most notably, Newell, Jaffee and Stavins (1999) show that between 1960 and
1980, the typical air-conditioner sold at Sears
energy-efficient.
On
became
significantly cheaper, but not
the other hand, between 1980 and 1990, there was
little
much more
change in costs,
but air-conditioners became much more energy-efficient, which, they argue, was a response
to higher energy prices.
In a related study,
Popp
(2002) finds a strong correlation between
aggregate patents and energy prices. These findings are consistent with the hypothesis that the
type of innovation responds to profit incentives, though they do not establish causality.
Third, there
is
substantial research focusing on innovation in the pharmaceutical industry.
Henderson and Cockburn (1996), Cockburn and Henderson (2001), and Danzon, Nichelson and
Sousa Pereira (2003) study the determinants of success
firm and project
size.
Galambos and Sturchio
in clinical trials, focusing
mainly on
Henderson and Stern (1999),
(1998), Cockburn,
Gambardella (2000), and Malerba and Orsenigo (2000) discuss various aspects of the recent
technological developments in the pharmaceutical industry.
investigate the complementarity between
spread of
Most
new
new
Ling, Berndt and Frank (2003)
technologies and the skills of physicians in the
drugs.
closely related to this study are Lichtenberg
and Waldfogel (2003) and Finkelstein
(2003). Lichtenberg
and Waldfogel show that following the Orphan Drug Act there were larger
declines in mortality
among
individuals with rare diseases (compared to other diseases) because
of the incentives created by the Act to develop drugs for these rare diseases. Finkelstein exploits
three different policy changes affecting the reimbursement of costs of vaccination against 6
infectious diseases: the 1991 policy change that all infants be vaccinated against hepatitis B, the
1993 decision of Medicare to cover the costs of influenza vaccinations, and the 1986 introduction
of funds to insure vaccine manufactures against product liability lawsuits for vaccines against
polio, diphteria, tetanus, measles,
mumps,
rubella, or pertussis.
She
finds that increases in
vaccine profitability resulting from these policy changes are associated with a significant increase
in the
number
The
of clinical trials to develop
rest of the
paper
is
new
vaccines against the relevant diseases.
organized as follows.
We
outline a simple
5
model linking innovation
to market size in the next section. Section 3 briefly explains our empirical strategy, and Section
5
Lichtenberg (2003) also presents evidence suggesting that the types of
useful for the elderly after Medicare was established.
more
new drugs changed towards drugs
4 describes the basic data sources and the construction of the key variables. Section 5 presents
the empirical results and a variety of robustness checks. Section 6 contains some concluding
remarks, while the Appendix gives further data details.
2
We now
Theory
framework to analyze the influence of market
outline a simple
size
on innovation.
Subsection 2.1 discusses the basic model. Subsection 2.2 derives the implications of potential
delays in the development and approval processes of
basic
model to
future market
new
drugs. Subsection 2.3 generalizes the
investigate the response of innovation effort
R&D to anticipated changes in
and
Finally subsection 2.4 extends the model to introduce an explicit distinction
size.
between entry of generic drugs and non-generic drugs.
Basic
2.1
Model
Consider an economy consisting of a set / individuals, each denoted by
t
6
[0,
and
bo),
First, a basic
all
y,
which can be consumed or used
for research expenditure.
is
Individual
(t),
....,
(t).
qj
is
continuous
production of other goods, or
,
xj.
....,
Each drug has a
yi (£) at
time
t.
potentially time-
Each individual demands only one type of drug. Hence, we
J
partition the set I of individuals into
if i
for the
has an exogenously given endowment
i
a large number J of drugs, x x
varying "quality", q t
such that
Time
individuals are infinitely lived. There are two types of goods in this economy.
good,
Second, there
i.
G Gj then individual
,
i
disjoint groups, Gi,...,Gj
demands drug
j
.
More
with G\
specifically,
U G^ U
if i
...
U Gj =
€ Gj then
,
/,
his
preferences are given by:
/
exp (-rt) [a
1
(i)
"7
(qj (t) Xji
(t)Y] dt,
(1)
Jo
where r
Ci (t) is
is
the discount rate of the consumers (also the interest rate in the economy), 7 G
the consumption of individual
of individual
i
of drug j.
of substitution equal to
6
1
6
i
of the basic
good
at time
t,
and
Xji (t) is
(0, 1),
the consumption
This Cobb-Douglas functional form, which implies an elasticity
between the basic good and drugs, and the assumption that each
Alternatively, suppose the preferences of individual
i
are:
j
f
exp (-w
1-7
(t,
qj (t) Xji (£))) c t (t)
dt,
Jo
where
uj (£, qj
(t)xji (£))
individual consumes
rt
— 7 In
is
{qj (t)xji (£)),
the discount rate, which
is
influenced by the quality and quantity of drugs that the
because they extend his life or affect
we obtain the expression in (1) in the
(e.g.,
its
quality). If
text.
we assume that u
(t,
qj (£) Xji (i))
=
individual only consumes one type of drug are for simplicity and do not affect the
Normalizing the price of the basic good to
j at time
t
by
Pj
(t),
individual
X
for all
6 I and
i
At any point
for all j
=
1,
for drugs are given
^ t)=
/
,.,
^#
P
(£),
is
FDA) and can
We
final
drug of
we think
for
this type. Equivalently,
(the rate of entry of
new
by
e Gj
lo s
(2)
there
If
can produce one unit of drug with quality
line j
that any
is
an innovation
new drug
new
drug
if
line j leads to
total
is
(t)
=
SjZj
some
lines
than others, which
is
(for
For
1.
example by the
indefinitely).
technology whereby one unit of the
t is
Sj
Zj (£),
>
of discovering a
new
the flow rate of innovation
is
(t)
(3)
.
Differences in Sj's introduce the possibility that technological progress
difficult in
where A >
(t)
approved
a flow rate of
R&D effort at time
nj
is
with
line j currently
under patent protection
R&D
drugs) for this line of drugs
drug
for
of quality Xqj
innovation
start with a very simple formulation of the
R&D
price of drug
brae!,
be sold to consumers immediately (and
good devoted to
7
one firm with the best-practice technology for producing each
this leads to the discovery of a
the purposes of the model,
i
and denoting the
results.
J.
using one unit of the basic good.
quality qj
for
f
{
type of drug. The best-practice firm in drug
qj (t)
periods,
demands
...,
in time, there
1 in all
main
is
scientifically
more
the effect emphasized by science-driven theories of
innovation discussed in the Introduction.
The most important
progress
is
feature of this
R&D technology for our focus here is that technological
directed in the sense that firms can devote their research effort
and expenditure to
developing particular types of drugs. This contrasts with a different model where firms invest
in
R&D
in
an undirected way, and discover new versions of any one of a
pharmaceutical industry, especially in recent past,
companies with
fairly sophisticated
research for specific drug lines
7
(e.g.,
R&D
is
set of drugs.
The
a prime example of an industry where
divisions or specialized
R&D
firms can undertake
Gambardella, 2000, Malerba and Orsenigo, 2000)
8
One
implication of the Cobb-Douglas functional form is that the share of income that an individual spends
is constant. This implication can be easily relaxed by considering a utility function with an elasticity
of substitution different from 1, as in the factor market models with directed technical change (see, for example,
on medicine
Acemoglu, 1998, 2002).
It is also straightforward to extend the model so that each individual demands potentially more than one
type of drug, though this would require additional notation.
8
Naturally, there exist examples of research directed at a specific drug type leading to the discovery of a
different product, such as the well-known example of Viagra, which resulted from research on hypertension and
angina, and was partly accidentally discovered from the detection of side effects in a clinical study (see, e.g.,
Kling, 1998).
The demand curves
would
j
is
like to
have an
in (2)
elasticity equal to 1, so
an unconstrained monopolist
charge an arbitrarily high price. However, the firm with the best drug in line
competing with the next best drug in that
charging price pj
(t).
If this price is arbitrarily high,
market and make positive
profits, driving
Consider such a firm with quality
line.
the next best quality could supply to the
the best technology to zero profits. Therefore, the
firm with the best drug has to set a limit price to exclude the next best firm
that consumers are happy to buy from
rather than
it
next best firm charges the lowest possible price,
and
if
(2),
qj (t)
then her instantaneous
marginal cost,
its
and price Pj
t
7
A^fe(*)) (l-7)
limit price equalizes these
1
if
the
Suppose that
1.
and chooses her optimal
(t)
time
utility at
she will have
1,
to ensure
i.e.,
will be:
she purchases from the next best firm, which, by definition, has quality
charges price equal to marginal cost,
The
—
buy from the next best firm even
equal to
i.e.,
a consumer buys from the best firm with quality
consumption as given by
qj (t)
qj (t)
/A and
utility:
^77 W(*)-
two expressions, hence, equilibrium
prices for all j
and
t
satisfy:
'
Pj(t)
The
profits of the firm
(4)
where the second
line defines Yj (t)
demanding drug
j.
Here
=
]T]
Yj
all
of demographics-driven changes in
(t)'s
t
are:
ieG yi
.
are
(5)
(t)
known
income of the group
as the total
market
in advance,
of
consumers
size (total sales) for
which
drug
j.
plausible in the context
is
demand. They can change because the number of consumers
product changes, or because their incomes change, or also because new varieties
from
of drugs steal consumers
independent from quality,
this particular drug. Notice that profits of
qj (£)
,
which
is
a feature of the Cobb- Douglas
Firms are forward-looking, and discount future
of profits for firms can be written
r Vj
(*
I
Qj)
-
Vj
The discounted
by a standard dynamic programming
(t
|
qj)
=
ttj
( 9i )
-
SjZj
(t
\
drug companies are
utility function.
profits at the rate r.
the value of a firm that owns the most advanced drug of quality
9
j at time
(X-l)jYj(t)
X'yYj (t) corresponds to the
Throughout we assume that
qj (t) in line
= (A-l) 7 $>(*)
=
this
X.
with the best product of quality
nj( Qj(t))
demanding
=
recursion.
qj in line j at time
qj ) Vj
(t
|
Vj
(t
|
t, is:
qj )
Throughout, we assume that the relevant trans versality conditions hold and discounted values are
value
qj),
9
(6)
finite.
—
for each j
is
1, 2,
equilibrium
where
J,
...,
R&D
time
effort at
the flow profits in drug line j given by
(qj) is
7Tj
on
t
this line
(because of the standard replacement effect
the best technology does not undertake any
To
1992).
we
simplify notation,
=
rij (t)
5jZj
(t)
R&D itself,
new
(t
(t
(t
|
innovation, thus the current firm will lose
(t)
and
subsection 2.4,
There
some drug
research for
profits
we present a model with separate entry
R&D
line j
=
we might have
Alternatively,
profitable,
for
with Vj
An
and
at time
0,
t,
(t)\
given by
equilibrium
is
is
>
0.
in
and non-generics.
if
there
is
positive
\
qj)
<
1.
(7)
in
1,
which case research
=
1, 2,
...,
so that Zj
(t)
is
not
be no innovation.
(t)\._
1
j that satisfy (4),
R&D levels
Zj (t)\
=1
consumer
j that satisfy
(6).
we must always have
Vj
(t
qj)
=
for
=
each j
1, 2,
Substituting this equation and (7) into (6) yields the levels of
Yj (t)'s are such that
t,
(t
=
qj)
that satisfy (2) and sequences of
i€l
.At)-™
and
rates of generics
sequences of prices pj
unique equilibrium:
each j
(t
|
\
for
and non-generic drugs, and
straightforward to characterize. Differentiating equation (7) with respect
to time implies that
Zj (t)
we think
then the free entry condition ensuring zero
then 5jVj
and 5jVj
0,
in equilibrium, there will
drugs x t
(•)
>
=
Zj (t)
equilibrium in this economy
demands
(7)
J
its
must hold. In other words,
if Zj (t)
An
market. Thus for now,
to develop better quality drugs. Therefore,
1, 2, ...,
Zj (t), especially
entrant are independent of the quality of
sufficient to take over (part of) the
free entry into
is
new
as corresponding to the entry of both generic
rij (t)
its
customers
steal
can also be included in
generics,
since equation (5) shows that the profits of a
of Zj
(qj)
but also takes into account that at the
qj),
\
equal to the flow profits, nj
qj), is
profits thereafter.
from the incumbent, such as the entry of
it is
is qj
Intuitively, the
qj).
Note that entry of new drugs that are not technologically superior, but
product, as long as
q3 )
\
example, Aghion and Howitt,
|
and make zero
best-practice status,
see, for
will typically use Zj (t) instead of Zj
there will be a
Zj (t
emphasized by Arrow, 1963, the firm with
flow value of owning the best technology in line j, rVj
flow rate
and
by other firms when current technology
first
plus the potential appreciation of the value, Vj
(5),
J,
all
=
and
for all
t.
-
1)
jYj
Prom now
(t)
R&D
J
as long as
effort in the
^-™®-' *},
(8)
on, unless otherwise stated,
equilibrium research levels are strictly positive,
(Sj (A
...,
-
r) /5j,
and we
will often
we assume that
i.e., Zj (t)
drop the
max
>
for all j
operator.
The most important
which
more
feature of (8)
is
The
the main focus of this paper.
is
profitable
that
highlights the market size effect in innovation,
it
greater
the market size for a particular drug, the
is
to be the supplier of that drug,
it is
Our
research effort to acquire this position.
empirical
and consequently, there
work below
will
be greater
investigates the strength of
this effect in the pharmaceutical industry over recent decades. In addition, naturally,
R&D as captured by 5j also increases R&D, and a higher interest
R&D since current R&D expenditures are rewarded by future revenues.
productivity of
Another important implication of
At any point
in time, the
amount of
determined by the current market
do not
this equation
which ensures that whenever there are
rate reduces
that there are no transitional dynamics.
devoted to developing a particular drug line
Past market sizes and anticipated future market
size.
affect current research effort.
R&D to
effort
is
a higher
This
is
an implication of the
(t
\
=
qf)
0.
The
sizes
technology,
immediately be
profit opportunities, there will
arbitrage them, thus ensuring Vj
R&D
linear
is
sufficient
intuition for the lack of response to
anticipated changes in future market size here highlights an important effect in quality ladder
models of technological progress: with a greater market
own the
best-practice product at the time
Investing in
However,
it
R&D
is
in
With the
linear
this innovation
model
Combining equations
(3)
too
much
and
(8) gives
nJ
larger
market
(t)
=
5J
market
entry of
is
drug).
approval of
new
size materializes.
little bit in
not profitable. 10
new drugs
as (ignoring the
max
operator):
(X-l) 1 YJ (t)-r.
(9)
This equation relates innovation or entry of new products, which we
FDA
some other firm
in advance, since
by the time the new and
in the size of the
will
approximate with
drugs, to market size (total expenditure of consumers in this line of
In addition, this equation also encapsulates the alternative view of the determinants
of innovation, which maintains that cross-drug distribution of
R&D
is
determined largely by
technological research opportunities or perhaps by other non-profit related motives.
are large
and potentially time- varying
differences in <5/s, then these
determining variations in
R&D across drug lines,
Whether
is
10
like to
achieves this objective.
it
can change discontinuously, so investing even a
here, Zj
advance of the actual increase
R&D
would
the market size has actually become larger.
advance could be useful to the extent that
not beneficial to invest in
would improve over
when
size in the future, firms
or not this
is
so
and market
size
may be
may
If
there
the primary factor
have only a small
effect.
an empirical question.
In practice, companies may also have an incentive not to market their discoveries before the market size
increases in order to prevent competitors from leapfrogging their new product.
Delays
2.2
The
of
model ignores the potential delays
baseline
new drugs
may be
Development and Approval
in
(for
much
as
example, DiMasi et
development and approval
in the process of
1991, report that the eventual marketing of a drug
al.,
To incorporate such
as 15 years after the beginning of initial research).
in the simplest possible way, suppose that
it
takes an interval of length
T
after the research
decision for the drug to be developed, gain approval, and enter the market. Therefore,
have Uj
(t)
Given
=
5 3 Zj
—
(t
T), where
this structure, the
rVj
(t
gj )
|
Equation (10)
is
a delayed
so a general analysis
is
Zj (t
— T)
the rate of innovation at time
key value function changes
-
Vj
(t
|
q3 )
differential
more
is
difficult.
=
ttj
fe)
-
(t-T)>
which recognizes that innovation
0,
t
—
we now
T.
to:
SjZj
(t
- T)
Vj
it
|
(10)
q3 )
equation rather than an ordinary differential equation,
Nevertheless, the unique equilibrium in this case
easy to characterize because of the simple structure here.
if Zj
delays
Now
then exp (-rT) SjVj
effort at
(t
\
time
t
—T
the free entry condition
=
q3 )
will lead to
is still
is:
(11)
1,
revenues at time
t,
hence the
discounting for the interval of length T. Equations (10) and (11) together imply that:
^-T)=^l&t!me^m&zi.
i0
which
is
very similar to
(8),
except for the term exp (— rT).
makes
it
clear that longer
(12)
This term takes into account
that because of the development and approval delays, costs of
benefits accrue. This equation also
\
R&D
are incurred before the
development and approval delays
discourage innovation.
Equation (12)
may
to future market sizes.
not the actual
R&D
give the impression that there should
This
is
now be a response
we measure
not the case, however, since what
expenditure, but entry of
prediction of the model for entry of
new
drugs. Since
new drugs changes
n3
-
(t)
=
5jZj
(t
of innovation
in the
—
data
is
T), the key
to:
nj (t)=exV (-rT)6j (\-l) 1Yj (t)-r,
which only
in
differs
from
(9)
by the term exp (—rT). This analysis therefore shows that delays
development and approval processes do not change the basic predictions relating to entry of
new
drugs, though the predictions about the timing of
this is useful to
R&D
bear in mind when we look at patents below).
and patenting are
different
(and
Anticipation Effects
2.3
We now generalize this
We change the baseline
R&D
for
is
basic setup to obtain a reaction to (anticipated) future market sizes.
model
one dimension: we assume that one unit of
in
in line j leads to the discovery of a better
drug at the flow rate
new drug
the aggregate research effort devoted to the discovery of a
$ (z) <
assume that
for all z,
which implies that greater research
returns within a given period (there are constant returns to scale
throughout
zcf)
cj)
good spent
5jZj(fi (zj),
where
in this line.
effort
when
final
We
Zj
also
runs into decreasing
(z)
=
but
1 for all z),
that greater aggregate research effort always
(z) is strictly increasing in z, so
leads to faster innovation in the aggregate.
R&D
Finally, free entry into
for all lines implies that
any new firm can enter taking
as given, thus without taking into account the reduction that
rates of other firms.
rVj
11
Given
(t
|
for
—
each j
SjZj
(t)
(/)
1, 2,
...,
J,
-
q3 )
entry causes in the innovation
changes
this specification, the value function
Vj
(t
q3 )
|
which only
rather than SjZj
(zj (t))
its
=
differs
tt,-
(q3 )
from
-
(6)
6jZj
(t)
Zj (t)) Vj (t
(
Zj (£)
|
to:
q3 )
(13)
,
because the flow rate of innovation
is
now
(t).
Since each potential entrant takes the aggregate research effort in each line of drug as given,
it
anticipates that one unit of the basic
innovation at the flow rate
5$ (zj
(£)).
good spent
=
An
1, 2,
...,
equilibrium
Zj (£)lj=i
j
To make
J
is
(again as long as Zj
>
(t)
nave to satisfy (14) instead of
and rearranging, we obtain J equilibrium
(zj (t))
— —$ [zj
t
r
[Zj it))
+6
^
(*)
now
assume that Yj
(13)
^
line j will lead to
an
is
(14)
with Vj () given by
(7)
remain constant over time. Differentiating
W^ = e^
r f!^
drug
0).
further progress in this case, let us
Zj [t)
in
=1,
defined similarly to before, except that
sizes
where e$
R&D
Thus, the free entry condition
5 j <t>{z] {t))VJ {t\q3 )
for each j
for
(t)
the sequence of
R&D
levels
(13).
=
Yj for
(14) with respect to time,
all t,
so that
market
and substituting into
differential equations in Zj (£)'s:
& W) - W (^
(t)) Zj (t) /4> (zj (£)) is
(*))
(^
the elasticity of the
4>
-
1)
lY
3
\
(15)
function.
11
This is the natural assumption given free entry. The alternative would be to assume that there is a
consortium of firms in each line, jointly maximizing profits. In this case, the free entry condition below would
change to: 5j [</> [z3 (qj (t))) + z3 (q3 (t))
(z3 (q3 (*)))] V3 {q3 (t)) = 1. This does not affect any of the results of
<j>'
the analysis.
10
now assumed
Since consumer incomes are
have
Zj (£)
=
be constant, the steady-state equilibrium must
to
R&D
This implies that the steady-state
0.
drug
level in
line j
=
1, 2, ...,
J
is:
5j(f> (zf)
In the special case where
the steady-state
size increases
drug
line,
An
R&D levels
R&D,
=
(.)
this
a greater
Sj,
higher interest rates reduce
z?,
with the
Figure
1.
to (8) in the basic model. Moreover,
hand
drug
lines.
side of (15)
is
for this
R&D.
that the equilibrium behavior of
is
profitability in other
substantially. In addition, the right
=
down
which corresponds to better research opportunities
important feature of this equation
pendent of research and
Zj (t)
equation boils
have the same features as the equilibrium there: a greater market
R&D, and
increases
1,
Zj (t)
is
inde-
This feature simplifies the dynamics
whenever
strictly increasing in Zj (t)
which implies that there can be at most one intersection of the right hand side
axis,
and at
this point of intersection, Zj (t) JZj (t) is increasing in Zj (£), as
drawn
in
Therefore, (15) defines an unstable differential equation. This implies that starting
away from the
steady-state, the equilibrium Zj
(t)
has to immediately
jump
to
its
steady-state
value as given by (16). Hence, there are no transitional dynamics in this extended model either.
However, there
now an
is
equilibrium response to anticipated future changes in market
Consider the following situation:
some
Yj in
equilibrium
jump up
future date
R&D?
t
>
be changing before
t,
t
=
in Zj
(i),
in particular
small amount initially at t
equilibrium.
t.
is
will this fully-anticipated
no change
But
in particular, Vj
the free entry condition, (14).
jumps
How
t'.
Suppose there
discontinuously at
suddenly announced at date
it is
=
in Zj (t) until
t.
t'
that Yj will increase to
change in market
This implies that
this implies that anticipating this
(t
\
qj)
<
0.
Since
Zj (t) is
jump,
R&D
size affect
Zj (i)
Vj
(t
\
has to
qj) will
constant, this would violate
This reasoning implies that there should be no anticipated
no jump
t',
at
t
=
t,
which
is
only possible
if Zj (t)
jumps by a
and then smoothly increases towards the new steady-state
Therefore, in this model there are no transitional dynamics starting
the steady-state, but
size.
away from
responds to anticipated future market size changes. Nevertheless,
the same considerations as in the baseline model limit this response. In other words, even
a change in market
size
is
anticipated far in advance, increasing
advance would not be profitable because another firm
is
likely to
R&D
investment too far in
innovate further before the
actual increase in market size materializes. In terms of our empirical work, even
changes are anticipated at least 20 or 30 years in advance, we
responses
12
A
much
later,
if
may expect
if
demographic
significant innovation
perhaps 5 or 10 years in advance or even contemporaneously 12
previous version of the paper also investigated a number of other generalizations, which we omit because
11
Generics and Non-Generics
2.4
The
analysis so far did not distinguish between generics
work, we
first
look at
and then distinguish between generics and non-generic drugs
all entries,
which better correspond to "innovation"
model change when we incorporate a
this in the simplest possible way,
discover
new drugs
and non-generics. In the empirical
We now
.
distinction
how
briefly discuss
the predictions of the
and assume that pharmaceutical firms can engage
a generic drug to the market does not involve
original research,
it still
resources spent upfront (for the approval process or for marketing).
Let us suppose that one unit of the
drug
at time
for
is
given by gj
drug line j at time
(£)
t.
non-generics influences 6j
must be
If
Ojhj
there
is
will
some degree
(£),
time
where
t
if
Although bringing
requires substantial
14
hj
at the flow rate 6j
(£)
is
.
We
assume that
new
9j
>
5j
,
total generic
for the
market
new
generics
so entry of
,
development expenditure
since introducing a generic to the
market
drug.
entry of a generic into drug line j at time
1/2). Recall that
then they
R&D to
In practice, in addition to technological factors, length of patents on
easier than inventing a
and the generic entrant
[0,
=
13
good devoted to prepare a generic
final
line j leads to successful entry at
t
in
do
as described above, or they can engage in costly development to introduce a
generic version of an already-existing drug (without quality improvement).
in
We
between generics and non-generics.
receive profits of
fj.
(A
—
1)
jYj
t,
(t)
we assume that both the incumbent
where Yj
(£) is
defined in
(5),
and
/j.
6
the generic entrant and the incumbent engage in Bertrand competition,
both charge marginal
cost,
and we would have //
=
0.
The formulation here
of non-Bertrand (e.g., Cournot) competition or collusion, so
[i
>
is
allows
possible. If
of space constraints.
Briefly, we allowed research to be only imperfectly directed, so that research towards the drug line j leads to
a flow rate of pSj of discovering a new drug of this type, and to a flow rate of (1 — p) Sj>/J any j' = 1, ..., J,
where 1 > p > 0. When p = 1, we have the model of subsection 2.1, while with p —> 0, research becomes
undirected. Interestingly, this generalization does not affect the results of the model, and even with p — > 0, the
innovation rates are given exactly by (9), and thus the results do not depend on the assumption of perfectly
directed research.
We also generalized to set up by allowing "technological drift" meaning random innovations arising from nonprofit and non-economic motives, which again did not affect the results, and led to an equilibrium relationship
similar to (9). Details are available upon request.
,
13
An alternative approach would be to link the entry of non-generics directly to the delayed entry of generics,
since generics can only enter after the patents on previous non-generics expire. Nevertheless, because introducing
a generic into the market still involves significant upfront costs, we expect profit incentives to play a major role
here.
hybrid model incorporating both such delays and profit incentives is significantly more complicated to
analyze.
A
14
Prior to 1984, the FDA demanded a similar application process for approval of a generic drug as for a
non-generic. For example, the generic drug's safety and efficacy had to be demonstrated through clinical trials.
DiMasi et al (1991) suggest that the cost of approval for a generic was as much as 40% of the cost of a nongeneric. The Hatch- Waxman Act in 1984 allowed a company to obtain FDA approval for a generic drug by
demonstrating that it is molecularly similar to and has the same active ingredients, indications and strength as
the non-generic, substantially reducing the costs of introducing a generic. Reiffen and Ward (2002) estimate the
cost of introducing a new generic to be
dollars before (DiMasi et al, 1991).
around $5 million
12
after the Act,
compared
to over one
hundred million
[x
=
0,
there would be no entry of generics, and the results in subsection 2.1 would apply.
In addition,
we assume
producer, there
that in a market that already includes the incumbent and a generic
no further room
is
a third producer, so we can ignore the potential entry
for
from further generic producers. Allowing
for further entry of this sort introduces additional
notation, but does not affect our qualitative results.
Given
rVj
(t
|
where
this structure, the value of innovation (for
q3 )
iTj
-V
d
{q3 )
=
(t\ q3
/i
—
(A
n3 fo) - 8jZj
=
)
l)jYj
as before
(t)
producers supplying drug j at time
above
is
(q3
)
V
(t
3
W
3
The main
t.
(t
3
h3
q3 )
\
(t)
is
difference
(t
|
qj),
[V3
(t
given by
W
-
q3 )
\
between
monopoly
its
is
3
(t
|
-
q3 )
W
3
is
(17)
,
(t
\
(t)
and
(6)
there will be
and the associated
q3 ).
With a
similar
given by:
W
(t
3
q3 )
|
=
^
(q3
)
-
53 z3
(t)
(t
3
|
(t
3
because there
q3 ))
\
this expression
position,
W
Intuitively, the only reason why the flow of profits captured by W
rW
(t
3
the value of being one of two
and becomes one of two producers, receiving value
reasoning to before, this value
is
now
is
the last term, which takes into account that at the flow rate 93 h3
entry of a generic, in which case the innovator loses
value Vj
-
q3 )
|
and
a non-generic)
\
q3 )
(18)
.
qj) will
come
to an end
a better drug introduced to the market, which happens at the rate 53 z3
(t).
Free entry requires that
Zj (t)
hj
{t)
>
>
and
0,
0,
and
These conditions are similar to
53
V
63
W
{t
3
(7)
(t
3
<
q3 )
\
\
qj)
with complementary slackness
1
<
1
with complementary slackness.
above, but written out explicitly in the form of a comple-
mentary slackness condition to emphasize that there may not be
R&D
for
new drugs
or
any
generic entry under certain conditions. Let us next assume that
fj,9j
for all j.
to
become
If this
<
Sj
(19)
assumption does not hold, entry of new generics
new
profitable for pharmaceutical companies to introduce
quality improvements
come
to
an end. As long as Assumption
to before imply that the unique equilibrium
*(t)
=
hj{t)
=
is
is
so rapid that
drugs,
it
ceases
and consequently,
(19) holds, similar
arguments
given by:
^(*-lW(0-r
(2Q)
t
(*j-^i)
Qj
13
(A
-
-1)7*5(0
.
6j
can also be
It
verified that if
\x
= 0,
(19) does not hold, there will
Prlimt ^oo
hj
(£)
=
the equilibrium of Proposition
be no
R&D,
0) as there will eventually
i.e.,
Zj (t)
=
in the
and
rij (t)
market
Assumption
if
(t)
=
drug
j.
hj
for
(or
gj (£), are:
= ^(A-l) 7 ^(t)-r,
{t)
9j{5j-^j)
0j
(A
~
(21)
-l)iYj{t)
Sj
There are a number of important points to note about
rates of both non-generics
and
and thus limf _>oo
be two producers
Moreover, the entry rate of non-generics and generics,
n3
0,
1 applies,
this equilibrium.
and generics respond positively to market
model generalizes the key prediction of our baseline model
size.
First, the
entry
Therefore, this
in subsection 2.1.
Second, other
comparative static results are now quite different than in the baseline model. The entry rates
of non-generics no longer respond to 5j (and Zj
positively to
profitable)
/j.
and
This
.
(£) is
decreasing in 5j). Instead, they respond
because the rate of entry of non-generics
is
make
(two parameters that should intuitively
6j
is
entry of generics more
determined to ensure
for generics; once generic drugs enter the market, their producers will continue to
until there
to Yj
(it
(£)
is
a new and better drug.
and
size
— 8j
6j
to be large,
than non-generic
fi
or (5j
and
entry.
—
fi9j) / [6j
—
5j) is larger). Plausibly,
therefore, generic entry to
r
—
>
0,
we can take
logs
on both
log
where
rrij (£)
=
XjYj
(t) is
drugs (or innovation),
tion) in
\i
size
to be small
be more responsive to changes
in
market
rij (£)
sides of equation (9) to obtain:
=
constant
+ log 5j + log rrij
the market size for drug line j at time
rij (£),
as
new drug
approvals by the
t,
16
(22)
(£)
FDA
t.
We
measure entry of new
(Food and Drug Administra-
This measure, denoted by
N
ct
for
drug
includes entry of generic drugs. Although generic drugs do not correspond
to "innovation" according to the standard use of this term, they are
16
we expect
market
Empirical Strategy
broad drug categories as described below.
category c at time
15
in
Empirical Specification and Estimation Issues
3.1
also
profits
15
3
As
make
Finally, differentiating the equations in (21) with respect
shows that either generics or non-generics may respond more to changes
depends on whether
profits
still
driven by the
same
In practice, the presence of uncertain delays in the development and approval process for non-generics
imply a smaller response to the current market size for these drugs.
Besides
new drug
rates: clinical trials.
may
approvals and patents, which we also use below, there is a third proxy for innovation
were unable to obtain data on clinical trials for a sufficient number of drug categories.
We
14
and are similar to innovation
profit incentives as innovation,
After presenting results using
Instead of actual market
we
size, m,j (t),
Md
demographic changes, which we denote by
Adding other potential determinants, time
we
general preferences than Cobb-Douglas,
where
Na
X'ct
size,
is
number
the
is
a
coefficients, C c s are
e ct
is
/i 's
t
=
ct
new drugs
0.
take a
number
log
where
equals
N =N
ct
1
ct
when
if
More
is
is
would with more
(23)
t
t,
M
ct
is
potential market
as the corresponding vector of
fi
common
time component, and
finally,
omitted influences. The specification with the
all
ensures that drug category fixed effects and
it
>
first
where iVct
=a
1
and
log
a count variable (number of new drugs), so
is
ct
is
equal to
Ma +X'
ct
N =
ct
1
if
i.e.,
This makes
0.
-/3
we change
ct
0,
=
if
ct
(23) to
1
N =
ct
(24)
t
and the variable dct
N =
and d ct
should be treated).
ct
N
ct
,
so
it
=
is
a
dummy
otherwise.
that
This
Hall,
and
The drawkback
is
that the
can introduce various biases.
to consider the following Poisson
Hausman,
impossible to estimate
+ j-d + ( c +n + e cU
N =
dct
it
used by Pakes and Griliches (1980), has the advantage of simplicity
mechanically a function of
is
N
that
not common, but in most specifications there are typically
data determine how
satisfactory
1999, 2002, or
model
(see, for
example, Wooldridge,
Griliches, 1984):
Na = exp(a
17
it
+ C c + V + e cU
of approaches to this problem. First,
ct
ct
is
there are no approvals,
flexibility (the
variable d ct
cells
N
N
procedure, which was
and
as
1
an estimating equation of the form:
in category c in time period
useful, since
is
(23)
In our data, this
a few drug category-time
We
construction below.
its
to differ from
ct
-f3
ct
by
have proportional impacts on entry of new drugs.
can equal
(23).
+X'
ct
a random disturbance term, capturing
effects
M
are a full set of time effects capturing any
One problem with equation
it
M
size driven
of category fixed effects that correspond to the \og5j terms
full se t
dependent variable in logarithm
time
a- log
market
and an error term capturing other unob-
effects
arrive at
the two types
differs for
will use potential
a vector of controls, including a constant, with
'
above,
of
N
and entry
and discuss
,
served influences, and allowing the coefficient of log
log
our model. 17
drug approvals, we separate generics from non-generics, and
all
investigate whether the relationship between market size
of drugs.
in the context of
•
log
M
ct
+X'
ct
-P
+ ( C + ^) + e
ct
,
(25)
generics and non-generics are imperfect substitutes, generics would correspond to new products,
would increase consumer welfare in a manner similar to technological improvements, and in fact,
formally correspond to technological change in the product variety models (e.g., Romer, 1990, Grossman and
Helpman, 1991).
and
Moreover,
if
their entry
15
which
is
a slight variant on equation (23) above. Estimates from the two models are typically
when
similar
there are only a few
when we look
empty approval
When
we
more empty
estimation of (25) would lead to biased estimates, however, since the fixed
and Griliches
(1984),
and transform
—
To
N
ct
/
Y^ T =i ^ct
is
the
in the sample.
we
deal with this problem,
.logMJ + ;C,.g + ft
number
follow
effects,
Hausman,
the
Hall,
)
of drugs approved in category c at time
the total number of drugs approved in category
T
and
c,
is
the total
number
divided by
t,
of time periods
This transformation removes the drug category dummies, and the coefficient
of interest, a, can be estimated consistently.
squares (NLLS) as well as
maximum
We
estimate this equation using nonlinear least
Woodridge (1999) shows that NLLS
likelihood (ML).
estimation strategy has good consistency properties, even
we
as
(25) to obtain:
ex P ( a
_
where Sct
cells,
favor the Poisson model.
C c 's, cannot be estimated consistently.
Finally,
there are
separately at generics and non-generics, there are larger discrepancies between
the two models, and in those cases
The
cells.
also report
when the
true model
is
not Poisson.
some estimates from the negative binomial model which
distributional assumptions of the Poisson
model
maximum
in a
relaxes the
likelihood context (see, for
example, Wooldridge, 2002, or Hausman, Hall, and Griliches, 1984).
In addition to equations (24) and (26), which have the contemporaneous value of
we
on the right hand
side,
whether there are
significant delays
reported by DiMasi et
from the time of
al.
also estimate equations with leads
(1991),
initial research.
and anticipation
it
effects.
and
Such
can take as long as 15 years
lags of log
M
ct
logMci
to determine
effects are possible, since, as
for
a drug to enter the market
Furthermore, changes in demographics can be anticipated
a long time in advance, so drug approvals
may respond
to anticipated future market sizes, as
highlighted by our analysis in subsection 2.3.
3.2
Potential Market Size and Identification
Throughout, we exploit the potentially exogenous component of market
size driven
by demo-
graphic trends, combined with differences in the age distribution of users for different types of
drugs.
data,
We obtain the age composition of users
and the changes
Our measure
in U.S.
(and expenditure) from micro drug consumption
demographics from the
of potential market size
CPS
(Current Population Survey) data.
is:
M
ct
= ^2u ca
a
16
p at
,
(27)
where p at
is
the U.S. population or income of age group a at time
measure of consumption of drug category
c
by individuals
in age
t,
and u ca
group
is
the time-invariant
We take time periods
a.
to be either five-year or ten-year intervals.
We
at time
compute
t
for
M
ct
in
two alternative ways:
p at and the average number
we
first,
use the U.S. population for age group a
of drugs in category c used per person in age group
a for u ca and second, we use the total income of age group a for p at and the average share of
;
drugs in category c in the total expenditure of those in age group a for u ca
we
refer to
The
latter,
be as the income-based measure, corresponds more closely to market
theoretical model, which
will
.
size in the
a combination of the number of consumers and their incomes, and
is
be our main measure.
which
18
For both measures, the over-time source of variation
changes in individual use, but purely from demographic changes captured by p at
—
is
i.e.,
not from
u ca 's are
not time-varying. So for example, changes in prices, which potentially result from innovations
and
affect
consumption patterns,
The major
will
threat to the validity of our empirical strategy
omitted variables (any variable that
effects).
not cause over-time variation in
not time-varying
is
Omitted variables related to market
is
is
M
ct
.
from potentially time-varying
taken out by the drug category fixed
may
size or profit opportunities
induce a bias in
the implied magnitudes, but will not lead to spurious positive estimates of the effect of market
size (in other words, the presence of
If
is
More threatening
of the appropriate market size).
omitted supply-side variables.
such variables
our instrument
We
in supply-side determinants of entry
is
essentially equivalent to
mismeasurement
to our identification strategy would be
valid,
it
should be orthogonal to variation
attempt to substantiate our identifying assumption
further by including lagged dependent variables, and adding controls such as pre-existing trends
and proxies
for other incentives to
field.
19
Data and Descriptive Statistics
4
4.1
undertake research in a particular
Basic Data Sources
The demographic data come from
0-20, 20-30, 30-50, 50-60,
these age groups.
March CPS, 1964-2000.
the
and 60+.
We
construct five age groups,
These divisions are motivated by drug use patterns of
To construct income
shares,
we
divide household income equally
among
the
If preferences have a Cobb-Douglas form as in (1) and are stable, the expenditure measure of u ca will always
be constant. With non-Cobb-Douglas preferences, changes in prices will induce changes in u ca over time. In
this case, by using a time-invariant measure of u ca we remove this potentially endogenous source of variation.
19
Another source of endogeneity may be that innovations in certain drug categories extend the lives of the
elderly, thus increasing their
Lichtenberg (2003) provides evidence that new drugs extend lives. This
ct
source of endogeneity is not likely to be quantitatively important, however, since the variation resulting from
extended lives in response to new drugs is a small fraction of the total variation in
Nevertheless, we will
ct
also report estimates that instrument
ct with past demographics, purging it from changes in longevity.
,
M
.
M
M
17
.
members
of the household. Figure 2 shows population shares for the five age groups,
3 shows the corresponding income shares
by
total income).
show a
large
trace the
To
amount
facilitate
(i.e.,
income of the corresponding age group divided
comparison with Figure
4, this figure starts in
1970.
of variation across age groups over time. In particular,
baby boomers,
and Figure
Both
it is
figures
possible to
as the fraction of those in the age bracket 20-30 in the 1970s,
and
those in the age bracket 30-50 in the 1980s and the 1990s.
FDA
The
classifies all prescription
subdivided into 159 categories.
intent
and chemical structure. 20
drugs into 20 major drug categories, which are further
These categories are based on a combination of therapeutic
We
drop 4 of the 20 major categories from
Anesthetics, Antidotes, Radiopharmaceuticals and Miscellaneous.
We
21
categories by breaking 10 of the 16 remaining categories into finer groups
this classification:
obtain a total of 34
when
there are signif-
icant heterogeneity in terms of the age distribution of users across subcategories.
We
separate
the major categories of Antimicrobials, Psychopharmacologics, Nutrients, Hormones, Dermatologies, Neorologics,
Ophthalmics, Otologics, Pain Relief and Respiratory because within these
categories there are subcategories with significantly different age profiles of users. For example,
within Antimicrobials, 0-20 year-olds use Antibiotics (except Tetracyclines) the most, while
Antivirals are used
most by people 30 and
Our main data source
is
for
drug use
is
older.
Appendix Table Al
lists
the 34 categories.
the Medical Expenditure Panel Survey (MEPS), which
a sample of U.S. households over the years 1996-1998. The survey has age and income data for
each household member, and covers about 25,000 individuals in each year. There
is
and the amount spent on drugs. In
also
a
list
of
prescription drugs used by each person
(if
about 500,000 medications prescribed.
We construct drug use per person and expenditure share
for
each category and each of our
and shows a
large
amount
five
any),
age groups. Appendix Table
of variation across drug categories.
Al
all,
there are
reports these numbers,
Many
of the categories are
used more by older people than by younger, but there are numerous exceptions. For example,
Contraceptives are used most by 20-30 and 30-50 year-olds, and Penicillins and Antibacterials
are used
market
most by individuals
in the youngest group.
size according to equation (27)
We
construct the measures of potential
by combining data from the
MEPS
and the CPS.
We supplement the MEPS data with the National Ambulatory Medical Care Survey (NAMCS),
20
Other authors, for example Lichtenberg (2003), use a different classification system based on diseases. For
our purposes, the FDA classification is attractive, both because it enables us to construct a better classification
of all approved drugs currently on the market, and because the microdata surveys give the FDA classes for the
drugs that the respondents use (see Data Appendix for details).
2
'We drop the Anesthetics, Radiopharmaceuticals and Miscellaneous categories because most of the items
in these categories were not developed for a distinct market. Radiopharmaceuticals are used for diagnostic
purposes, and the Miscellaneous category mainly contains surgical and dental tools. The Antidote category is
dropped because there are few drugs approved and there is little use of these drugs in the surveys. See the Data
Appendix for further details on the contruction of our categories.
18
which
an annual survey of doctors working
is
in private practices.
The survey
includes drug
use for the years 1980, 1981, 1985, and 1989-2000. Observations are at the doctor-patient-visit
level;
Doctors are selected randomly, surveyed for a
there are about 40,000 visits per year.
week, with patient-visits selected randomly from the week.
that
it
The main use
of the
NAMCS
is
covers a longer time period, enabling us to check whether the age composition of users
across categories has changed over time. Using the
categorization, with 30 categories.
Table
1
NAMCS
data,
we
construct a second drug
22
gives correlations between various measures of drug use. Panel
NAMCS
of correlation between the
A shows a high degree
surveys at various dates, indicating that the age profile of
users has not changed significantly over the 1980s
and the
The
1990s.
overall correlation
row
We also report weighted correlations, where observations
MEPS or NAMCS. Weighted correlations are larger than the
looks at the unconditional correlation.
are weighted by cell size in the
overall correlation because our estimates of the age distribution of users are less precise
there are fewer individuals using drugs in a particular category.
correlation
third
row reports mean
by drug, which calculates the within category correlation between the two measures
and then averages
it
across
all
categories. This
whether the age distribution of users
Panel
The
when
for
measure
is
more informative
for the question of
a particular drug has changed over time.
B performs the same calculation for the three waves of the MEPS,
and similarly shows
C shows a
MEPS data.
substantial persistence in the age distribution of drug expenditure. Finally, panel
high degree of correlation between expenditure shares and use per person in the
But perhaps
is
surprisingly, there
is
low correlation between the
NAMCS
and the MEPS. This
because the two surveys yield very different estimates for total use of each category (but
very similar estimates of relative use by age groups within each category, as shown by the high
level of
mean
correlation
by drug).
We
conjecture that this reflects the fact that
which samples doctors in private practice rather than individuals,
the
is
NAMCS,
not as representative as
MEPS.
The
last
major data source
is
a
list
drugs, the so-called orphan drugs, 23
of
new FDA drug approvals. We exclude over-the-counter
and drugs that have the same identifying
characteristics
2
Appendix Table A2, which is available upon request, compares this classification system with the 34 category
system developed with the MEPS. Some of the 159 categories have been dropped from one classification system,
but not from the other, because there were not sufficient observations to construct reliable estimates of drug
use from one of the surveys. Appendix Table A2 shows that the two systems are closely related. In general
the 34 category system is a slightly less aggregated version of the 30 category system. Nevertheless, there are
a number of cases where a given FDA category is combined with a second FDA category in one system, but
with a third FDA category in the other system (e.g., Misc. Antibacterials are with Sulfonamides in the MEPS
system, but with Antiseptics in the NAMCS system).
23
These drugs treat rare conditions, affecting fewer than 200,000 people. An example is botox, first developed
to treat adult dystonia, which causes involuntary muscle contractions. We drop these drugs because we have
difficulty matching them consistently, and because they receive special inducements under the Orphan Drug
19
same name, company, and
(i.e.,
category, or the
the time period 1970-2000. Since
listed
we can
FDA approval number). We focus on
FDA categories for drugs that are still
same
only match
by the FDA, the quality of the approvals data deteriorates
addition, because
we
we go back
as
in time.
In
are using age composition from the 1990s, the quality of our measures of
potential market size also deteriorates as
we go back
in time.
Our approvals dataset
for 1970-
2000 comprises 6,595 prescription drugs, including both generics and non-generics (see the
Appendix). Since 1970 there have been about half as
many approved
non-generics as generics.
Figure 4 shows the share of drug approvals over time to compare with changes in income
shares depicted in Figure 3 (or population shares
we compute drug approvals
34 categories into
five
shown
in Figure 2).
To
over five-year intervals for the 34 categories.
construct Figure
We
4,
then combine the
groups, based on the age group that accounts for the largest fraction of
use in that category (thus this cut of the data uses only part of information that we will exploit
in the regression analysis)
and compute the share
,
of drug approvals
by dividing the number of
approvals in a given category by total approvals in that time period. 24 Comparing this figure
with Figure
3,
a positive association between contemporaneous changes in population share
and changes
in
drug approvals
for the
corresponding age group can be detected visually. For
example, the income share of the 30-50 group increases over the sample, and so does the entry
of drugs most used by this group.
Both the shares
of
income and entry of drugs
on the other hand, show a downward trend, while those
for the
for those 0-20,
60+ age group show an
increase
followed by a decline. Table 2 also shows changes in income shares by age group and in
drug approvals
(for all
FDA
drugs as well as separately for generics and non-generics), and confirms
the patterns depicted in Figures 3 and
4.
These patterns are explored
in greater detail in the
regression analysis below.
5
5.1
Results
Basic Specifications
Table
3A
provides the basic results from the estimation of (24) and (26) with non-linear least-
squares (NLLS) and ordinary least-squares (OLS), and Table
estimates of the Poisson and negative binomial models.
measure constructed using
MEPS data.
3B
reports
maximum
likelihood
We start with the potential market size
These basic specifications do not contain any covariates
Act.
There are large fluctuations in the total number of approvals, presumably because of a number of institutional changes. For example, it was discovered in 1989 that some
officials were taking bribes to speed up
the approval process for generic drugs. As a result, in the early 1990s the approval process for generics was
greatly slowed. See, for example, The Washington Post, August 16, 1989. When we separate our approval data
into generics and non-generics, we see a large drop in generics approvals in the early 1990s, but only a small
decline for non-generics.
thank Ernie Berndt for suggestions on this issue.
FDA
We
20
other than drug category effects and time
of (potential) market size (log
In column
of Table 3A,
1
M
ct )
we
effects.
The
results
show a
large
and
significant effect
on entry of new drugs.
start with our basic "income- weighted"
constructed using expenditure data from the
MEPS
data
set,
measure of
logMcf
,
and income from the CPS, with
the time periods corresponding to five-year intervals. Since our estimates of age composition in
smaller categories are less precise, observations are weighted by
ture for income-based measures in the
The standard
1.95,
which
and
total
drug use
for
cell sizes (total
of
a
Huber- White
with a standard error of
in this basic specification is 6.04
The OLS
significant at the 1 percent level.
is
estimate
with standard error 2.00, thus significant only at the 5 percent
expendi-
population-based measures).
errors throughout are corrected for heteroscedasticity using the
The NLLS estimate
formula.
cell
MEPS
is
level.
somewhat
smaller, 4.52,
20
Potential market size appears to explain a sizable fraction of the total variation in the entry
new drugs
of
relationship
(for
is
example, the partial
of the
OLS
specification
is
0.20),
and the empirical
not driven by outliers. Figure 5 shows a plot of the residuals of log
the residuals of log
dummies
R2
M
ct
in the
OLS
regression, in
are taken out. Observations are labeled
both cases
by
their
after
N
ct
against
drug category and period
drug category codes
(see
Appendix
Table Al), and each code appears more than once, since there are multiple periods. The
the figure corresponds to the estimated relationship reported in column
1.
Although categories
43 and 61 (Anorexiants, Central Nervous System drugs and Vitamins/Minerals) typically
42,
the pattern less well and are outliers in either direction, excluding these has
fit
line in
our estimate of a. For example, dropping these three categories leads to an
little effect
NLLS
on
estimate of
7.07 (s.e.= 2.29).
The
a
1
quantitative magnitude of the effect in column
1 is large
but plausible, implying that
per cent increase in our market size measure leads to about a 6 percent increase in drug
approvals. Since there are a total of 6,595 approvals between 1970
and 2000, thus on average
32 approvals in every five-year interval in each of our 34 categories, the estimate of 6.04 implies
that a
1
percent increase in market size leads to the entry of about two
new
drugs. Total phar-
maceutical sales was approximately $130 billion in 1999 (IMS, 2000), which implies an average
annual expenditure of $3.8
A
billion per category.
1
percent increase therefore corresponds to
$38 million, or about $570 million over 15 years, which
costs for non-generics are
around $800 million
(in
is
the
life
of a typical drug. Since entry
2000 dollars, DiMasi
et al, 2001), while for
generics they have varied between $5 million and over $100 million (DiMasi et
and Ward, 2002), entry
25
of
two new drugs
in response to
al,
1991, Reiffen
an increase of $570 million
Clustering the standard errors at the level of drug categories has little effect because there
see the lagged dependent variable specifications reported in Table 6.
serial correlation
—
21
is
in
revenue
no residual
is
within the range of plausible responses.
In column
2,
we
use ten-year intervals instead of the five-year intervals, both to reduce
the potential noise in the entry of
new drugs during
the five-year intervals and also to check
whether ten-year intervals correspond more closely to the relevant match between market
and
the
entry.
1
The estimate
percent
The
level.
of
a
using
smaller
NLLS
NLLS
is
4.95 with standard error 1.90, significant at
estimate indicates that there might be some attenuation
with the ten-year invtervals, but the ten-year
estimate. In the remainder,
now
we mainly
OLS
estimate
is
slightly larger
also look at the effects of changes in
size
market
seems more
estimated with about the same level of precision.
sions.
without weights
is
The
reflects
illustrated
and the OLS estimates,
NLLS and OLS, and
are
equation (26).
number
We
in
columns
1
also smaller than in
conjecture that this
and
2,
but
columns
somewhat weaker
1
3B
in Table
l.
27
are broadly similar, and
methods. In panel
The number below
in curly brackets
is
estimates are unweighted. In the
The
and
relationship
A of that
the estimate
is
show that the main
table,
we use maximum
results are robust to
likelihood to estimate
the maximum-likelihood standard error, while
the robust standard error that does not impose the Poisson
structure to calculate the standard errors, and instead uses the Huber- White formula.
expenditure.
still
the less precise estimation of age composition in the smaller categories,
by the correlations
results in Table
different estimation
the
by population changes.
26
Without weights, the NLLS estimates are smaller than
are significant at the 5 percent level.
which
columns 3
see the effect of weighting on the estimates, columns 5 and 6 report unweighted regres-
significant at the 1 percent level,
2,
satisfactory, in
size driven purely
Using this measure leads to estimates that are slightly larger for both
To
than the five-year
focus on five-year intervals.
Although the income-weighted measure of market
and 4 we
size
first
two columns, we compute market
size using
The
income and
estimates are larger than the corresponding estimates in columns 5 and 6
of Table 3A. In panel
B we
use a weighted
maximum
likelihood procedure,
estimates to those in columns 1-4 of Table 3A. Finally,
we
and obtain similar
also report estimates
from the
negative binomial model in panel C, which allows for overdispersion of the Poisson parameter.
In this case, the estimates of
a
are slightly smaller than those in columns 1-4 of Table 3A,
and
26
We also estimated these models using West European and Japanese demographic information in addition
to U.S. information (obtained from the United Nations website, esa.un.org/unpp/). Using the total populationbased market size measure combining European, Japanese and U.S. populations gives similar results to those
obtained using only U.S. information. For example, the NLLS specification using this total market size and
five-year intervals gives an estimate of 5.27 (s.e.=1.50), as compared to the corresponding estimate of 6.16 in
column
3.
27
This conjecture also receives support from the fact that when we construct our potential market size measure
using the NAMCS, which has a more even distribution of observations across drug categories, the unweighted
and weighted results are similar (see Table 7 below).
22
still
significant at 1 or 5 percent.
Delays and Anticipation Effects
5.2
The
and approval processes are
theoretical analysis suggested that delays in the development
unlikely to create delays in the entry of
there
is
room
for
new drugs to enter
new drugs
market
in response to changes in
size,
but
before the actual increase in market size because of anticipa-
tion effects, especially since demographics-driven changes in market size should be anticipated
We
in advance.
including lags
investigate the role of delays
(logMCit -i) and
leads
(logMc>t+1 )
and anticipation
of potential
subsection by
effects in this
market
size
on the
right
hand
side
of our estimating equations.
Columns
Table
3A
1
and 2 of Table 4
for comparison.
replicate the basic five-year
Columns 3 and 4 include the market
and show that new drug entry responds mainly
on
lag market size
market
is
Columns
5
For example, in column
and 6 show that
lag
market
size
A in column
significant delays in the response of
In column
7,
we
where
5,
level.
The
new drug
1).
it is
These
is
and previous
the coefficient on current market
significant but
on lead market
size is
much
results therefore
sizes are individually insignificant,
coefficient
when entered by
much
itself
smaller than the
show no evidence
market
size.
size.
Both the contem-
but jointly significant at the
larger.
of
In column
8,
when we
1
use
same magnitude, but current market
significant at the 1 percent level (and similar to the baseline) while the lead
market
when we
include
insignificant.
only future market
effects,
3,
entries to changes in
ten-year intervals, instead, the coefficients are about the
size
size (in fact, the coefficient
include the current and one period ahead market
poraneous and lead market
is
from the previous period
also typically insignificant
is
estimate with current market size in column
size
market
from
than our baseline, 13.56 (s.e.=4.45) with NLLS, and 8.62 (s.e.= 3.91) with OLS. 28
(with the exception of Panel
percent
to current
size
specifications
negative, though insignificant, presumably because current
sizes are highly correlated).
size is larger
and ten-year
Columns 9 and 10 show
size.
perhaps with
29
five-
These
large
significant coefficients
results therefore suggest that there
or ten-year leads, which
anticipation effects highlighted
and
by the
is
may be some
anticipation
consistent with the possibility of limited
theoretical model.
30
28
We construct the lagged market size measures for 1960s using demographic information from the CPS, so
the number of observations does not decline. The results are similar if we only use the post-1970 data.
29
We have also extended the sample for the lead specification by combining the population projections of
the U.S. Census Bureau for 2010 with the incomes from the 1990s (the results are also similar if we use a
linear extrapolation to predict future income). Using this additional period yields similar results. For example,
re-estimating the specification in column 4 with NLLS, the estimate on current market size is 3.10 (s.e.=3.75)
and the estimate on lead market size is 4.33 (s.e.=4.03), jointly significant at 1 percent. The specification in
column 6 gives an estimate on lead market size of 7.10 (s.e.=1.86).
30
Here "limited" does not refer to the strength of the effect, but to the fact that the response to market size
23
Potential Supply-Side Determinants of Innovation
5.3
we
In this subsection,
investigate the robustness of the baseline results to controlling for po-
tential non-profit determinants of innovation, such as changes in scientific incentives or oppor-
by the
tunities captured
First, recall that the
permanent
<S/s in
the theoretical model.
major threat to our
identification strategy
differences in 5/s are already taken out
5/s change over time, they are also
our basic specifications
Columns
specification,
1
and
is
likely to
by our drug category
serially correlated.
log
OLS
ct
iff-
log
N
ct
-i
+j-d + 5 c + n + e
to
ct
t
ct
(28)
.
correlated with the error term mechanically, estimates to this equation would
this
problem by instrumenting log
is
the arguments for the exogeneity of
The estimates
The
of
a from
coefficient
N
ct -\
logMct
is
more generally
also useful
new drugs
equation (28), reported in columns
on the lagged dependent
is
no
effect
a
(see, for
to check for
1
and
2,
variable, logA^ct _i,
though
are quite similar to
is
essentially
significant with
is
OLS,
and
it is
no significant residual
either changes in scientific opportunities or other sources,
controlling for such serial correlation has
is
(such serial correlation would
again insignificant with NLLS. These results therefore show that there
plausible conjecture
This
.
less compelling).
insignificant for five-year intervals; for ten-year intervals,
due to
with logA^c£ _ 2
no additional autocorrelation in the error term, e ct
other sources of serial correlation in the entry rate of
A
\ogNct
to check for the importance of these concerns.
example, Blundell and Bond, 1998). This specification
serial correlation,
lags of
lag of the dependent variable, logiVct_i, to our basic
N = a- log Ma +
valid instrument as long as there
the baseline.
Adding
version, the basic regression therefore changes to:
be biased, and we deal with
make
fixed effects). If the
2 of Table 5 report the results of estimating a lagged dependent variable
specifications. In the
is
way
therefore a simple
by adding a one-period
Since log7Vct _ 1
be
changes in the 5/s (since
is
and that
on our estimates.
that non-profit incentives to develop drugs would be particularly
responsive to opportunities to save lives or cure major illnesses. Motivated by this reasoning,
our second strategy looks at variation in the health benefits of new drugs across categories.
New
drugs in our data set include both drugs that are demanded by the consumers but do not
"save lives", such as Prozac, Paxil, Vioxx, or Viagra, or those that actually save lives such as
heart medicines or cancer treatments (see Lichtenberg, 2003, on the effect of pharmaceutical
innovations on declines in mortality).
To
investigate this issue,
we measure
the
number
of
5-10 years before the change in market size, not further in advance. If we include further leads of market
these are much smaller and insignificant. For example, the 15 years lead when included by itself is highly
insignificant; the NLLS estimate is 1.57 (s.e.=2.52), and the OLS estimate is 1.16 (s.e.=6.06).
is
size,
24
corresponding to each drug category using the Mortality Detail Files from the
life-years lost
National Center for Health Statistics from 1970-1994. Following Lichtenberg (2003), for each
we subtract the
death,
person's age from 65, then calculate the total
for all the deaths resulting
We
add
from diseases related to drugs in each category.
measure of
this
models as a proxy
number
hand
life-years lost to the right
for this source of non-profit incentive to
of life-years lost
31
side of our baseline regression
undertake research. Since we are
using mortality data prior to 1995, we drop the last time period from the regression.
baseline regression for the years 1970-1994
an estimate of approximately the same
reported in column 3 of Table
is
size as the baseline using all years.
5,
and leads
Column
on the
life-years lost variable (unreported) is small
Starting in column
3,
in addition to
B, which use current market
size,
we
NLLS
and
estimates in panel
also report
NLLS
insignificant.
A and OLS
categories.
we
The
32
estimates in panel
estimates with leads of market size in
panel C. These specifications typically yield results very similar to those in Table
Third,
to
4 reports
the result of using the life-years lost variable, and finds no change in our estimate of a.
coefficient
The
4.
investigate the implications of differences in scientific funding for various drug
Using the Computer Retrieval of Information on
(details in the
Data Appendix) we construct a
,
funding for research projects in
all
Scientific Projects
variable measuring the total
drug categories, and include
(CRISP) dataset
amount
this variable as
of federal
a control on
the right hand side. To the extent that government funding also responds to potential market
example, because drug companies have a greater tendency to apply for funding in
size (for
areas where they plan to do research), this variable would be correlated with our market size
measure. In practice, the correlation
this variable, or
both
its
is
low,
and columns
current and lag values, has
5
and 6 show that the inclusion of
little effect
on our estimates of
a.
Fourth, to control for potential trends in scientific opportunities across drug categories,
we add
A =
We
proxies for pre-existing trends.
construct an estimate for pre-existing trends as
(logA^o — log7Vc40 )/2, where logA^o
c
log iVCv 4o
is
the log approvals in 1940.
log
Na = a
log
We
is
the log approvals for category c in 1960 and
then estimate the equation:
Ma + Y^
^c-(7i
+
5c
+
j2 t
+e
ct
,
(29)
i=80,90
where
ctj's
are
dummies
for the 1980s
and 1990s. This
specification allows
drug categories that
31
For example, if someone dies at age 32, this counts as 33 life years lost; people dying older than 65 receive
no weight in this calculation.
Note that the Mortality Detail Files are coded by disease class, so we must convert the classification to our
system. Since many of our categories contain diseases or conditions that do not lead to death, we obtain a
number
32
We
found
of
empty
cells.
also ran separate regressions using fiveno change in our estimates of a.
and ten-year
25
lags of
life
years lost (both unreported), and again
have grown at different rates between 1940 and 1960 to also grow at different rates
and the 1990s. Column 7 reports the
our baseline estimate, 6.23, with standard error
A =
c
(log7VC|70
—
log
-/VC]40
)
The estimate
results of this exercise.
Column
1.88.
8 repeats the
The
/3 as the measure of pre-existing trends.
of
in the 1980s
a. is
same
similar to
exercise with
resulting estimate
is
again similar to our baseline. These results are perhaps not surprising, since pre- 1970 approvals
are considerably noisier, thus only an imperfect control for pre-existing trends.
An
To do
alternative,
so,
we
and substantially more demanding, strategy
estimate:
log
where c
refers to the
category,
i.e.,
to include linear time trends.
is
N
ct
=
a-
log
M
ct
+
r]
t
c
+ 5 + \i + e
34 detailed drug categories, and
the one which detailed category c belongs
to be captured
C
C
t
ct
(30)
,
refers to the relevant 16
We expect
to.
major drug
technological differences
by which of the 16 major drug categories each drug belongs
to, since these
categories are based on therapeutic intent, while the subcategories are based on use
The
group.
estimates, reported in
NLLS, the estimate
Using lead market
a
of
is
is
save space), 4.99 (s.e.=2.81),
The OLS estimate
market
size results in
significant at 1 percent, while the
is
For example, with
are close to our baseline.
9,
5.60 (s.e.=2.10).
size instead of current
estimate, 8.46 (s.e.=2.53),
We
column
is
smaller and insignificant.
a similar pattern: the
OLS
by age
NLLS
estimate (not reported to
significant at 10 percent.
also investigate the potential effects of advances in biotechnology, such as the use of
recombinant
DNA,
and the 1990s. In
or other technological changes, during the late 1980s
terms of our model, these developments would correspond to changes in the
we drop
6j's.
the categories of Cancer and Cardiovascular, which, according to the
In column 10,
FDA approval list,
have witnessed the entry of the greatest number of orphan drugs, presumably by biotechnology
firms.
Although, as noted above, our dependent variable does not include these drugs, we also
check whether our results are driven by entry of
new drugs
in
column 10 are
a
are not sensitive to dropping these two categories.
close to those in
In addition, there
is
we drop
To
ucts
1 of
and Thyroid category) and
in the
first
active in produc-
Hematologic category. 33 In column
these two categories, and again find that our results are essentially unchanged.
assess the role of biotechnology firms further,
known
The estimates
Table 3A, demonstrating that the estimates of
anecdotal evidence that biotechnology firms were
ing insulin (the Glucose
11,
column
in these categories.
as biologies,
we add the approvals
where biotechnology firms have been
of a group of prod-
active, to our
measure of drug
Biotechnology firms were also active in producing human growth factor, but since there are only a small
of individuals using these drugs in the MEPS and NAMCS, these drugs are not included in our approvals
number
dataset.
26
approvals. These products include
some
vaccines, blood
and plasma related products, and other
products such as interferon and erythroproteins (used for red blood
cell
included in our baseline measure because they go through a separate
The
results of this regression, reported in
Finally, to see
column
12,
show
little
production), and are not
FDA regulatory
change in the estimates of a. u
whether the advent of biotechnology or other technological advances of the
past two decades have changed the relationship between market size and entry of
we estimated our
dummy
a
is
new
drugs,
baseline models including an interaction between a post-1985 (or post-1990)
and market
Our
size.
estimates showed no evidence of significant interactions.
example, in a specification parallel to the
of
process.
NLLS model
column
of
5.24 (s.e.=1.96), and the interaction with the post-1985
1
For
of Table 3A, the estimate
dummy is
0.10 (s.e.=0.07), thus
quantitatively very small and insignificant.
The
results in this subsection therefore
show that a number
new drugs have
market-size related) determinants of the entry of
findings.
Although these
results are not conclusive
of controls for other (non-
on the
little effect
on our main
effect of scientific or other non-profit
considerations in pharmaceutical research, they suggest that the effect of potential market size
on entry and innovation
Changes
5.4
Our market
trends.
in
size
is
relatively robust.
Health Insurance Coverage
measure only exploits changes in potential market
Another source of variation
in
market
size
size driven
comes from changes
by demographic
in coverage of
drug
expenditure in private or public health insurance programs. Finkelstein (2003), for example,
exploits changes in the coverage of various vaccines to estimate the effect of these policies
the development of
new
vaccines.
During our sample period, there were
in health insurance plans.
60+
significant changes in the coverage of
drug expenditure
For example, the percentage of 0-20 year-olds with some form
of private health insurance coverage
while the percentage of
on
fell
from about 73% to
68% between
year-olds with private insurance rose from
62%
1974 and 1996,
to
75%
(authors'
calculations from the National Health Interview Survey). Furthermore, there have been changes
in
Medicaid
eligibility rules,
designed to insure more poor children. These changes in health
insurance coverage induce additional changes in market
sizes.
We now investigate both whether
controlling for this source of variation affects the estimates of the impact of the demographics-
driven potential market size measure,
market
34
size
M
ct
,
and whether we can improve our measure of potential
by including information on health insurance coverage.
We
(12),
also repeated our basic regressions without the categories where biologies are most common, Antivirals
Hematologics (20) and Immunologics (80), with little effect on the estimates of a.
27
We
use the National Health Interview Survey (NHIS, 1974-1996) to construct the fraction
of each age group covered
by a private health insurance plan. Because there
is
no consistent
information on prescription drug coverage, we assign prescription coverage to any individual
with both doctor and surgical coverage. Prescription drug coverage
this
measure
in the years
where we can observe
The NHIS
it.
is
highly correlated with
also includes information
Medicaid and Medicare. Because Medicare does not cover prescription drugs,
perform a simple
enables us to
"falsification test"
with direct parallel to our
First,
it
on
M
ct
measure, we use the
NHIS
to construct the following
insurance coverage variable:
H
ct
= ^2u ca
-pa
-
fat
(31)
,
a
where fat
the fraction of age group a in period
is
expenditure share as described above, and p a
to isolate changes
add
Column
of
a
and
and weights)
.
We
use pa rather than p at
now
coefficient is
though
5.21, as
opposed to 6.04
NLLS
all
H
effect of log
adding the
H
ct
using Medicaid and Medicare coverage
individuals above 65 are covered by Medicare, not
ct
Columns
ct
is
all
measures
of
a
them take up
are unaffected
substantially weaker, presumably because these measures exploit
the variation in actual market
H
itself is
insignificant.
fat (though
less of
specification the
Table 3A. The coefficient on logj^
in
the benefits). Not surprisingly, with Medicaid and Medicare, the estimates of
much
then
results are robust to controlling for
For example, in the
Next, we construct alternative measures of log
and the
We
(26).
and that our baseline
separate trends in private health-care coverage.
rates for
the
of Table 6 shows that the addition of this variable has no effect on the estimates
1
relative to those in Table 3A,
positive,
size
due to drug insurance coverage rather than demographic changes.
logf/tf to our estimating equations (24)
is
the sample average income of age group a (in
is
we always use income-based market
this section,
with private health insurance, u ca
t
size.
in the specifications
7-9 take an alternative approach
both demographic changes and changes
Columns 4-6
with lead market
investigate the implications of
size,
and construct a market
and show similar
size
results.
measure incorporating
in insurance coverage:
H = ^V
ct
Q
p at
(32)
fat,
a
where u ca p a and fa are as defined above.
t
,
Using log
1
of Table
t
H
3A
ct
instead of log
M
ct
in
column 7 leads to a
(4.27 with standard error of 1.24).
28
The
similar coefficient relative to
difference in
column
magnitude may be because
insurance-driven and demographics-driven market sizes affect different types of drugs, or be-
cause the demographics-driven changes are anticipated further in advance, potentially enabling
a greater response. The most
market
size
We
than
M
likely explanation,
in
Medicaid
about
9%
1990s.
Column
eligibility
H^
that
is
H
ct
a worse measure of
is
use to those without insurance. 35
because (32) effectively assigns
ct
next construct an alternative measure of
Changes
however,
from information on Medicaid coverage.
have made children more likely to be covered. In the 1970s,
of 0-20 year olds were covered by Medicaid; this fraction rose to
the late
8 shows a very small and marginally significant effect of Medicaid insurance on
new
the rate of entry of
drugs.
Finally, as a falsification exercise,
Medicare take-up
not predict
17% by
rates. Since
new drug
we
use a measure of
H
calculated from information
ct
on
Medicare does not cover prescription drugs, this measure should
entries (though note that
M
ct
and
H
ct
are correlated by construction).
Reassuringly, the estimate in column 9 shows no positive effect of Medicare coverage.
Reverse Causality
5.5
Lichtenberg (2003) shows that
a lesser extent,
life
new drugs have increased
expectancy) by as
whereby the market
for reverse causality
because their users
much
live longer.
changes in population are
exploiting. Nevertheless,
We
likely to
we
the average age at death (and hence, to
as 1 percent per year. This introduces the potential
size for successful
think this
is
drugs
may be endogenously
not a first-order concern, since drug-induced
be small relative to the demographic changes that we are
further address this issue by instrumenting for current population
using the corresponding population from 10 years before. For example,
fraction of
50+
The
in 1980.
we
10 years later, but
is
50+
year-olds
unaffected by
is
highly correlated with the fraction of
new drugs
5.56
(s.e.
we instrument
size
with past market
1.
with NLLS, and both are significant at the
year-olds
causality. In
column
With NLLS, the estimate
is
results using ten-year intervals.
1
percent
level.
possibility is that the effect of demographics-driven market size was somewhat overestimated in
of the correlation between this measure and insurance coverage. However, the estimates in
1-6 in this table suggest that this is not a major concern.
3A because
columns
size.
Columns 3 and 4 show the corresponding
Another
Table
60+
estimates in column 4 give larger coefficients than the corresponding non-instrumented
results, particularly
35
market
year-olds
=1.69), slightly lower than the non-instrumented estimate for the same time period
reported in column
The IV
for
60+
that are developed in the intervening 10 years.
These instrumental- variables (IV) estimates show no evidence of reverse
2 of Table 7,
use the population
year-olds in 1970 as an instrument for the population fraction of
fraction of
larger,
29
Results from the
5.6
The
is
rest of
NAMCS
NAMCS
a useful exercise because the
categories than the
of drug
MEPS, and
NAMCS
as noted above, the
also because
private practice, whereas the
5
is less
MEPS
it
starts in 1980, enabling us to check
a sample of
all
MEPS
smaller than the corresponding
corresponding
MEPS
data with the
1
and
2,
since the
estimate, but
NAMCS
is
The OLS
estimate.
is
NAMCS
7,
data.
we use
results,
pattern from the
insignificant
estimate
is
also smaller
Column
is
NAMCS
results
than the
6 uses the
similar to
column 3
and Table 3A
is
not
probably driven by the non-representative
36
OLS
is
significant at the 5 percent level.
Columns
which are larger than the weighted estimates.
MEPS, and
is
8
5;
the
NLLS
and 9 report
This contrasts with the
consistent with our conjecture that the differences between
weighted and unweighted results in the
MEPS
were largely because of the
distribution estimates for the smaller categories in that data set (this
NAMCS,
in
and considerably
ten-year time intervals, and find similar results to column
insignificant while the
unweighted
is
and obtains estimates
classification system,
differences in classification systems, but
column
In
by doctors
NAMCS does not provide drug
significant at the 1 percent level.
of Table 3A. This shows that the disparity between the
nature of the
However,
biases.
U.S. households.
expenditure information). The estimate of a using NLLS, 2.86,
due to
whether use
shows our baseline regression, with the same specification as column 3 of Table 3A
(we cannot repeat the specifications of columns
MEPS
any
late 1990s introduces
representative, since the data are reported
is
This
has a more even distribution of users across drug
consumption and expenditure data from the
Column
NAMCS.
Table 7 repeats some of our main specifications using data from the
is
less precise
age
not a problem in the
which leads to a more precise estimate of age distribution of users
in the smaller
cells).
Finally,
we
use the
late 1990s induces
M^
980
36
,
NAMCS
to investigate whether our reliance on drug use data from the
any systematic
bias.
37
with the use per person numbers,
Using an alternative market
We
u^
construct an alternative estimate of market
80
,
only from the 1980
NAMCS
survey.
We
size,
then
constructed with the fraction of total use for each age group for
each age group, leads to very similar estimates from the two surveys, which
are also close to the MEPS estimates reported in Table 3A. Although this alternative measure has the advantage
of not being sensitive to the total expenditure by category in NAMCS, the baseline measure we use appears to
be both more natural and theoretically better motivated.
37
To see the potential reason for concern, suppose that a Gastrointestinal drug that is a major improvement
over existing drugs enters the market before the first year of the MEPS, 1996, and is consequently used by a
large number of 30-50 year-olds for the 1970-2000 period. The drug use and expenditure shares we estimate from
the MEPS for Gastrointestinals would include the use and expenditure of that drug. As a result of the entry of
this successful drug, we may overestimate the importance of Gastrointestinals for 30-50 year-olds. Nevertheless,
this will not lead to spurious positive estimates of the effect of potential market size on entry, since overtime
variation in our measure of market size is still purely driven by demographic changes.
u ca rather than use per person
,
size variable
for
30
estimate equations (24) and (26) in the post-1980 sample, instrumenting
logMrf
For comparison, using our baseline measure only for 1980 onwards with the
obtain a
NLLS
estimate of 4.64 (s.e.= 3.58) and an
logM^980
Instrumenting with
10.
3.98) with
NLLS and
in
OLS
column 11 leads
OLS,
6.63 (s.e.= 4.57) with
in
with log
NAMCS
M
1980
ct
(
data,
we
estimate of 5.76 (s.e.= 4.00) in column
to
an estimate of a equal to 4.79 (s.e.=
both cases similar to the estimate
in
column
10.
This comparison suggests that using microdata from the 1990s to construct rates of drug
use
is
unlikely to create any significant bias.
Generics vs Non-Generics
5.7
Table 8 shows the results of some of our specifications using only generic drugs to construct
and non-generics
separately, there are
year intervals, there are 9 empty
OLS
estimates (but
Column
1 in
we
more empty
cells,
respectively. Therefore, in Tables 8
NLLS
the
and
the
9,
2
(s.e. =2. 69), is
and 3
size,
results should
size, similar
to
column
is
be more
significant at the 1 percent level.
(s.e. =1.91),
is
reliable
investigate the robustness of this result
not significant and has
The estimate
than the
for five-year intervals
38
though
smaller,
right
hand
side
when we
The corresponding
still
significant at
is
more
little effect
control for serial corre-
and when we control
for
major drug
The lagged dependent
vari-
on the point estimates of a either with generics or
non-generics, though with generics, the estimate, 8.08 (s.e.= 5.34),
cent.
five-
in Table 3A. For generics,
1
category trends (respectively, estimating equations (28) and (30)).
able
drugs and
but the results also show a significant response of non-generics.
by including lagged entry on the
lation
all
larger estimate for generics suggests that the entry of generic drugs
responsive to market
Columns
NLLS
both Tables 8 and 9 reports the basic specifications
estimate, 7.70
The
example, using
also report the latter for completeness).
estimate in Table 9 for non-generics, 4.41
percent.
cells; for
but there are 27 and 24 for generics and non-generics,
using the income-based measure of market
1
Because we now look at generics
Table 9 reports similar regressions for non-generics.
Net.
is
for non-generics, 4.55 (s.e.= 1.91), continues to
only significant at 10 per-
be significant at
1
percent.
Controlling for linear trends, on the other hand, has almost no effect on the estimates, which
continue to be significant at
Columns 4 and
market
1
percent in both cases.
5 investigate anticipation effects. For generics,
sizes are entered together, neither
market
size
is
when both
significant,
current and lead
though the estimate on
38
Because the requirements to gain FDA approval for generic drugs changed in 1984 with the Hatch- Waxman
Act, we investigated whether the relationship between potential market size and new drug approvals also
changed. The answer seems to be no. For example, interacting market size with a post-1985 dummy in the
NLLS specification, the estimate of a is 7.94 (s.e. 2.71) and the coefficient on the interaction is 0.15 (s.e. 0.08),
i.e., quantitatively very small. There is a large increase in approvals of generics in the late 1980s followed by a
decrease in the early 1990s, which are mostly captured by the time dummies.
31
current market size
and
When
small.
similar to
is
column
lead market size
is
1,
while the estimate on lead market size
entered by
itself, it is
significant for five-year intervals,
thus there might be some limited response of generics to anticipated future market
pattern
is
somewhat
different for non-generics:
entered together with current market size and
Columns 6 and 7
lead market size
when used by
itself.
insignificant
is
a negative and insignificant
a statistically significant
coefficient.
effect
On
and
significant
is
in
column
The
7.
inconclusive as to whether lag market size matters for generic entry
generics regressions lag market size
market
size
The
and by
FDA
insignificant
evidence
40
in the U.S.).
generate more revenues.
more
and
therapies.
The
Research by the National Institute
priority or
To look
at the relationship
significant innovation,
new
its
own has
is
therefore
FDA
sig-
has a second
for
Health Care Man-
molecules (but not both) generate
between market
we construct a new measure
approvals for drugs that are designated to be both
very similar to the estimate in column
standard error (3.48).
Column
1
9,
we
find
fewer
,
which only includes
entities
and
priority drugs.
an estimate of
4.06,
which
7 looks at the impact of the five-year lead market size, and
is
significant at 10 percent. Therefore,
these estimates are similar to the estimates from columns 1 and
many
ct
molecules
and a potential mea-
size
N
new
but statistically insignificant because of the large
obtains a coefficient estimate of 4.71 (s.e.= 2.79), which
are considerably larger.
of
new molecular
In the basic specification reported in column 6 of Table
presence of
has
In contrast, in non-
revenue than other non-generics, but those that are both priority drugs and
sure of
size
specifications in Table 9 to save space.
available products
agement (2002) suggests that drugs labeled
is
market
whether or not a drug contains a new molecule (an active ingredient that has not
been marketed before
less
lag
current and
both when entered together with current
and we do not report these
improvement over
The
both when
has labeled some non-generics as priority drugs because they appear to offer
nificant clinical
classification,
itself,
is
When
the other hand, lag market size on
on entry of generics
sizes.
39
of Table 8 investigate the impact of lag market size.
lag market size are present together, current market size
negative
is
Most probably, the
new molecular
2,
but the standard errors
significantly larger standard errors result
entities
and
priority drugs
from the
than generics or non-generics
in our data (there are only a total of 190 such drugs in our 1970-2000 dataset).
39
We
also experimented with longer leads, and obtained similar results. Lack of a significant anticipation effect
for non-generics is somewhat surprising, especially given the evidence that there might be some anticipation
effects for generics. This might be because the non-generic results are less precise, making it difficult to detect
anticipation effects. Note also that when we look at priority drugs and new molecules in columns 6 and
market size has more predictive power than current market size, but is only significant at 10 percent.
We
7,
lead
expected lag market size to matter and generic entry to lag behind non-generics because of patent
protection for non-generics.
32
Patents
5.8
we
Finally,
pharmaceutical patents from
Thomson Derwent
company, we mapped these patents into our
of patents for chemicals into
amount
of noise, which
and with the help
FDA classification system.
41
obtained data on
of a specialist at this
However, the mapping
FDA categories is imperfect and necessarily introduces a significant
makes inference more
Firms typically apply
Inc.,
We
on patents.
investigate the effect of potential market size
for
difficult in this case.
a patent prior to the
about 5-10 years before the drug
is
approved.
42
clinical trial stage of
Given the
drug development, or
results so far,
we might expect
patents to respond to future demographic changes.
In columns 8 and 9 of Table
the
number
similar
market
insignificant.
patents.
44
we estimate equations
of patents in a particular drug category.
magnitude to the estimates
for lead
FDA
9,
43
First, this
The
and
(26), except that
estimates of
a
for non-generics (e.g., 3.12 for current
N
ct
are positive
market
size
is
now
and of
and 4.38
but because of the much larger standard errors, they are statistically
size),
There
(24)
may
may
be a number of reasons
simply
reflect
categories, especially bearing in
for the
much
larger standard errors with
the imperfect match between the patent data and the
mind the
potential use of certain chemical structures in
multiple drug lines. Second, the significant costs and uncertainty involved in the development
of
new molecules and
patentable products
intended for the 1990s
may be
may be
creating substantial attenuation
a drug
patented in the 1980s or 1990s, depending on delays in the
research process). Third, pharmaceutical companies
may respond more
the later stages of the research process than at the earlier stages.
more responsive to
(e.g.,
OECD demand than to U.S.
to profit incentives at
Finally, patents
demand. To investigate
may be
this last possibility,
we
looked at the relationship between changes in market size derived from European, Japanese and
U.S. demographic changes. In this case,
we
on patents that are broadly similar to the
results for non-generics
find estimates of the five-year lead of
5 percent (for example, 4.23 with standard error 1.82 with
1.28 with
patents,
OLS). Although
we
this result suggests that
are currently unable to
make more
OECD
and
NLLS and
market
size
statistically significant at
3.07 with standard error
demand may be more important
for
progress in distinguishing between these various
We could not use the data from the Hall-Jaffe-TYachtenberg patent dataset (see Jaffe and Trachtenberg,
2002) because we were unable to map their classification based on chemical structure to our drug categories.
42
The firm therefore loses a significant fraction of the life of the patent before it can begin marketing the
drug. Part of the Hatch- Waxman Act allowed pharmaceutical companies to apply to the FDA for an extension
of the life of their patents, if they could show that they lost marketing time while waiting for approval. The
maximum extension is 5 years, and depends, among other things, on the length of the initial FDA approval
process. Overall, companies have a maximum of 14 years of patent protection after FDA approval.
43
We obtain similar results with ten-year or fifteen-year leads of market size.
44
Finkelstein (2003) also finds weaker results for vaccine patents than for later stages of development.
41
33
explanations, and the weaker results for patents remain a puzzle.
Concluding Remarks
6
This paper investigates the response of entry of
changes in potential market
a
results indicate that
1
new drugs and pharmaceutical
by U.S.
size of users, driven
(or
OECD)
innovation to
demographic changes. Our
percent increase in the potential market size for a drug category leads
to approximately 4-6 percent growth in the entry of
new drugs approved by
the
FDA. This
response comes from the entry of both generics and non-generics, though the effect on generics
is
larger
and somewhat more robust.
This finding,
if
further proven to be robust, has important implications both for research
on the pharmaceutical industry, and
for the
endogenous growth and directed technical change
literatures. It provides evidence that, as conjectured
by these models,
R&D
and technological
change are directed towards more profitable areas. Furthermore, the magnitude of the
which
is
important for evaluating various theoretical predictions of these models,
is
effect,
substantial.
For example, directed technical change models suggest that the relative demand curves for
factors
can be upward, rather than downward, sloping
responds to a
tions (21)
and
1
if
the development of
percent increase in market size by more than
(22) in
Acemoglu, 2002)
1
new
technologies
percent (see, for example, equa-
—the corresponding number implied by our estimates
is
in the range of 4 to 6. Second, these findings imply that pharmaceutical research towards drugs
with relatively small markets
(2002). Building
on
may be
this premise,
limited,
which
a key premise of recent work by Kremer
Kremer suggests that there needs
incentives for developing drugs against malaria
We
is
and other
third- world diseases.
view this research as part of a broader investigation of the
on innovation. Evidence from a
the pharmaceuticals
single industry
may be more
and
at the
economy-wide
effects of profit incentives
nonrepresentative, for example because
research oriented than other industries.
investigating the response of innovation
industries
may be
to be selective government
and entry
of
new products
Future research
to market size both in specific
level is necessary to substantiate the results presented here.
34
References
7
"Why Do New
Acemoglu, Daron. 1998.
cal
Change and Wage
Technologies Complement Skills? Directed Techni-
Inequality." Quarterly Journal of Economics. 113, pp. 1055-1090.
"Directed Technical Change" Review of Economic Studies.
Acemoglu, Daron. 2002.
69,
pp. 781-810.
Acemoglu, Daron. 2003. "Labor- and Capital-Augmenting Technical Change." Journal of
European Economic Associations,
1,
pp. 1-37.
Aghion, Philippe and Peter Howitt. 1992. "A Model of Growth Through Creative Destruction" Econometrica, 60, pp. 323-351.
Aghion, Philippe and Peter Howitt. 1998. Endogenous Growth Theory.
Arrow, Kenneth
Economic
J.
Press.
"The Economic Implications of Learning-by-Doing," Review of
1962.
Studies, 29, pp. 155-173.
Blundell, Richard
in
MA: MIT
and Stephen Bond.
1998.
Dynamic Panel Data Models" Journal
Ceruzzi, Paul E. 2000.
A
"Initial
Conditions and
Moment
Restrictions
of Econometrics, 87, pp. 115-144.
History of Modern Computing.
Cockburn, Rebecca M. and Iain M. Henderson.
opment: Unpacking the Advantages of Size
in
2001.
MA: MIT
"Scale
Press.
and Scope
in
Drug Devel-
Pharmaceutical Research." Journal of Health
Economics, 20, pp. 1033-1057.
Danzon, Patricia M., Sean Nicholson and Nuno Sousa Pereira.
Pharmaceutical-Biotechnology
R&D: The
2003.
"Productivity in
Role of Experience in Alliances."
NBER
Working
Paper #9615.
DiMasi, Joseph A., Ronald
W. Hansen and
Louis Lasagna.
1991.
"Cost of Innovation in
the Pharmaceutical Industry." Journal of Health Economics, 10, pp. 107-142.
DiMasi, Joseph A., Ronald
Innovation:
New Estimates
of
W. Hansen and Henry
Drug Development
G. Grabowski.
2003.
"The Price of
Costs." Journal of Health Economics, 22, pp.
151-185.
Drandakis, E. and
Distribution"
Phelps.
Economic Journal,
Finkelstein,
Industry."
Edmund
Amy.
NBER
1965.
"A Model
of Induced Invention,
Growth and
76, pp. 823-840.
2003. "Health Policy and Technical Change: Evidence from the Vaccine
Working Paper #9460.
Galambos, Louis and Jeffrey Sturchio. 1998. "Pharmaceutical Firms and the Transition to
Biotechnology:
A
Study
in Strategic Innovation."
Business History Review, 72, pp. 250-278.
Gambardella, Alfonso. 2000. Science and Innovation: The U.S. Pharmaceutical Industry
During the 1980s, Cambridge University Press, Cambridge UK.
35
1957.
Griliches, Zvi.
"Hybrid Corn:
An
Exploration in the Economics of Technological
Change." Econometrica, 25, pp. 501-522.
Grossman, Gene and Elhanan Helpman. 1991. Innovation and Growth in
omy.
MA: MIT
Hausman,
the Global
Econ-
Press.
Jerry,
Bronwyn H.
Data with an Application
Hall and Zvi Griliches. 1984. "Econometric Models for Count
Patents-R&D Relationship." Econometrica,
ot the
52, pp. 909-938.
Hayami, Yujiro and Vernon Ruttan. 1970. "Factor Prices and Technical Change in Agricultural Development: the U.S.
and Japan, 1880-1960" Journal of
Political
Economy,
77. pp.
1115-1141.
Henderson, Rebecca and Iain M. Cockburn.
1996.
"Scale,
Scope and Spillovers:
The
Determinants of Research Productivity in Drug Discovery." Rand Journal of Economics, 27,
pp. 32-59.
IMS. 2000. "World Review" IMS Health, at www.ims-global.com/insight/report/market_growth/re
Jaffe,
Adam
and Manuel Trachtenberg and Rebecca Henderson. 2002. Patents,
and Innovations. MA:
MIT
Kennedy, Charles 1964.
Economic Journal,
Kling, Jim.
(2),
Citations,
Press..
"Induced Bias in Innovation and the Theory of Distribution"
74, pp. 541-547.
1998.
"From Hypertension to Angina to Viagra." Modern Drug Discovery,
1
pp. 31-8.
Kremer, Michael. 2002. "Pharmaceuticals and the Developing World" Journal of Economic
Perspectives. 16 (4), pp. 67-90.
Lichtenberg, Frank R. 2001. "The Allocation of Publicly Funded Biomedical Research," in
Medical Care Output and Productivity, Studies in Income and Wealth volume 63. Ernst Berndt
and David Cutler,
eds. University of
Lichtenberg, Frank R. 2002.
Chicago Press, 2001.
"The Effects of Medicare on Health Care Utilization and
Outcomes." Frontiers in Health Policy Research.
Lichtenberg, Frank R. 2002b.
5,
pp. 27-52.
"Sources of U.S. Longevity Increases, 1960-1997."
NBER
Working Paper #8755.
Lichtenberg, Frank R. 2003. "The Impact of
New Drug
Launches on Longevity: Evidence
from Longitudinal, Disease-level Data from 52 Countries, 1982-2001".
Columbia University
mimeo.
Lichtenberg, Frank R. and Suchin Virabhak 2002.
Progress, Longevity and Quality of Life: Drugs as
Paper #9351.
36
"Pharmaceutical Embodied Technical
Equipment
for
Your Health."
NBER
Working
Ling,
Davina C, Ernie Berndt, and Richard G. Prank. 2003. "General Purpose Technolo-
gies, Capital-Skill
Complementarity, and the Diffusion of New Psychotropic Medications among
Medicaid Populations"
MIT
mimeo.
Malerba, Franco and Luigi Orsenigo.
2002.
"Innovation and Market Structure in the
Dynamics of the Pharmacetutical Industry: Towards a History Friendly Model."
and Corporate Change. 11
(4),
Industrial
pp. 667-703.
National Institute for Health Care Management. 2002. "Changing Patterns of Pharmaceutical
Innovation" at www.nihcm.org/innovations.pdf.
Newell, Richard,
pothesis
Adam
Jaffee
and Robert
Stavins.
1999.
"The Induce Innovation Hy-
and Energy-Saving Technological Change." Quarterly Journal of Economics.
114, pp.
907-940.
Pakes, Ariel and Zvi Griliches. 1980. "Patents and
Economic
R&D at
the Firm Level:
A First Look."
Letters, 5, pp. 377-381.
Pakes, Ariel and
Mark Schankerman.
Research Intensity." pp. 209-232 in
R&D,
1984.
"An Explanation
into the Determinants of
Patents and Productivity edited by Zvi Griliches,
National Bureau of Economic Research, University of Chicago Press, Chicago 1984.
Popp, David. 2002 "Induced Innovation and Energy Prices" American Economic Review.
92, pp. 160-180.
Reiffen,
David and Michael R.Ward. 2002 "Generic Drug Industry Dynamics." University
of Texas at Arlington,
mimeo
Romer, Paul M. 1990. "Endogenous Technological Change." Journal of Political Economy,
98. S71-S102.
Rosenberg, Nathan. 1974. "Science, Invention and Economic Growth" Economic Journal.
84, pp. 90-108.
Samuelson, Paul. 1965. "A Theory of Induced Innovations Along Kennedy- Weisacker Lines"
Review of Economics and
Statistics, 47, pp. 444-464.
Scherer, Michael. 1984. Innovation
and Growth. MA:
MIT
Schmookler, Jacob. 1966. Invention and Economic Growth.
Scott Morton, Fiona.
1999.
press.
MA: Harvard
University Press.
"Entry Decisions in the Generic Drug Industry." The Rand
Journal, 30, pp. 421-440.
Wooldridge, Jeffrey M. 1999. "Distribution-free Estimation of Some Nonlinear Panel Data
Models." Journal of Econometrics, 90, pp. 77-97.
Wooldridge, Jeffrey M. 2002. Econometrica Analysis of Cross Section and Panel Data,
Press Cambridge.
37
MIT
8
Data Appendix
Prescription Drug Use and Expenditure, and Drug Categories from the
8.1
The
MEPS
an annual survey of randomly sampled households; we use the 1996, 1997 and
is
1998 surveys.
We obtain each person's
age, the
name
expenditure (there are multiple records for people
Over the 3
MEPS
of the prescription drug(s) used,
who
total
use more than one prescription drug).
we have about 500,000 drugs used and about 75,000
years,
and
Expenditure
people.
includes out-of-pocket expenses, as well as amounts paid by insurers, collected from surveys
sent to insurance companies.
We
begin with the 159 therapeutic categories, obtained from the FDA's National Drug
Code (NDC)
The names
Directory.
Appendix Table Al. The
FDA approved
NDC Directory contains
drugs currently on the market.
159 categories by matching
it
We
We
We
divide these
file
with the therapeutic category for most
assign each drug in the
in the
of
MEPS;
in the
MEPS
NDC
to one of the
file.
We
these are usually not
cannot
commonly
than 5 percent of the total drugs used.
less
calculate drug use
a
by national drug code with a drug
match about 10 percent of the drugs mentioned
used drugs, and make up
column
of these categories can be found in the second
and expenditure by ten-year age group, using the survey weights.
by the corresponding population and income
for that year
and age group,
as
estimated from the CPS, to obtain drug use and expenditure per person for each age group
and category. The
results are very similar
of use per person in the survey,
methodology because
MEPS
if
we
construct use per person as a weighted average
without using
i.e.,
CPS
information.
We
prefer the former
enables us to construct income-based measures without relying on
it
income data.
The
FDA
has assigned each of the 159 categories to one of 20 major therapeutic categories.
Within each major category, we separate minor categories when there
in the age structure of
system).
is
sufficient heterogeneity
drug expenditure (using drug use yields the identical
From Table Al,
it is
apparent that we separate categories
when
there
variation. For example, within Antimicrobials (categories 10-12) category 10
is
classification
is
considerable
used more by
0-20 year olds, category 11 has a steadily upward sloping age profile of users, and category 12
is
used approximately equally by individuals over age 30.
As noted
in the text,
we drop
maceuticals, and Miscellaneous.
four major categories:
We
sufficient observations to estimate
as our cutoff rule.
We
we aggregate the
50-60, 60+, since
drop several minor categories when there are not
reliable
age structure.
We
use about 1,500 observations
obtain this number from observing that only categories with more than
1,500 observations have fairly
Finally,
a
also
Anesthetics, Antidotes, Radiophar-
smooth age
initial
profiles.
ten-year age groups into 5 age groups, 0-20, 20-30, 30-50,
most of our 34 categories show sharp peaks
38
in
one of the 5 groups.
Prescription Drug Use and Drug Categories from the
8.2
NAMCS
The
1981, 1985,
and 1989-2000
u ca we aggregate
,
visits. It
from the
differs
all
MEPS
in several important ways. First,
(in constructing
years of
NAMCS
Second, the survey
data).
1993-2000, the
We use this
NAMCS
map between
85%
information to construct
of prescribed drugs; in
these high rates,
still
We
based on doctor-patient
a
list
of drugs prescribed, with a
maximum
we
in constructing
drug use per person with
on the doctor's primary
diagnosis, but
it is
drugs prescribed and the diagnosis.
FDA category for each prescribed medication. 45
a mapping of medications to FDA class, which we can use
provides the
Our worst
to assign drugs from earlier survey years.
were
is
we do
also contains information
not possible to create a consistent
about
is
depending on the year. Since we do not have information on expenditure, we weight
MEPS). The survey
From
covers the years 1980,
does not cover doctors at institutions or hospitals, unless they are considered to have a
multiple drugs for a single patient equally (as
the
it
our classification system and estimating drug use,
private practice. For each visit in the sample, there
of 5 to 8,
NAMCS
success rate
most years the match rate
is
is
in 1980,
where we matched
well above
90%. Because of
believe that the bias from only using prescribed drugs in earlier years that
being prescribed in the early 1990s
initially construct
is
not large.
drug use per person by ten-year age group as we do with the
MEPS.
We obtain 30 drug categories, as shown in Appendix Table A2, which is available upon request.
We find that the same 5 age groups are suitable for the NAMCS data. We use the same
cutoff rule of 1500 observations for the NAMCS as for the MEPS for dropping FDA categories.
Different FDA categories are excluded from the two surveys, e.g., Antifungals from the NAMCS
and Anterior Pituitary from the MEPS.
8.3
Drug Approvals from the FDA
We have obtained a list of FDA drug approvals from Frank Lichtenberg.
and orphan drugs
(of
Over-the-counter drugs
which only a few can be matched) are excluded. Moreover,
biologies,
which
go through a separate approval process, are not in this dataset.
We
match drugs
in the approval
list
to
FDA
categories
by drug name.
13,916 of 16,220
prescription drugs (86%) approved since 1970 are matched, while before 1970, the
is
match
rate
about 45%. This motivates our focus on drug approvals between 1970 and 2000. Drugs that
have the same approval number and
which the corresponding
MEPS
FDA
are excluded. Finally,
FDA
category
is
class as a previously
dropped because of
approved drug and drugs
for
insufficient observations in the
we drop drugs with the same name,
MEPS
category and different
dosage from a previously approved drug, leaving us with our sample of 6,595 drugs. Of these,
1,928 are non-generics,
and 190 are
priority drugs
and new molecules.
45
Several drugs change FDA classes over the 8 years. In most cases,
these drugs stay within the same category; we drop those that do not.
39
when we construct the 30
categories
CRISP Data
8.4
The Computer
erally
Retrieval of Information on Scientific Projects (CRISP) dataset contains fed-
funded research projects at
universities, hospitals
and other
Many
institutions.
of the
grants are for very basic research, so they cover the earliest stages of drug development.
by the National
projects are funded
Institutes of Health
and the Substance Abuse and Mental
Health Services Administration, as well as a variety of other agencies such as the
Center for Disease Control and Prevention.
CRISP database
for
a total of about 11
Each record
projects
listed,
list
We
The
have obtained a dataset with
all
FDA
and the
pojects in the
1972-1995 from Frank Lichtenberg. In 1995 there were 57,553 grants, for
billion dollars.
lists
the project's investigator and
and the amount awarded. Most
affiliation,
one or several diseases which the researchers intend to study. For each disease
we know whether
it is
of primary, secondary or tertiary importance in the project. In
our analysis, we use only primary diseases, and divide the award amount evenly between them.
Our
results are not sensitive to alternative weighting schemes.
The
disease classification system has about 2,900 diseases, though these are arranged in a
heirarchical structure into 35
major disease
classes.
We map the detailed disease classes into our
classification
scheme, though there are no matches of primary disease to the
MEPS-based
categories, Contraceptives, Skeletal
ness, Non-narcotic Analgesics,
five of
the smallest
Muscle Hyperactivity, Vertigo/Motion Sick-
and Central Pain Syndromes. Our
results are not sensitive to
dropping these categories from the regressions.
8.5
We
Patents
have obtained patent data from Thomson Derwent
granted in the United States, between 1970-2000.
We
Inc.
We
use
Thomson Derwent
specialist at the
is
we
are not able to
into our
for
FDA
pharmaceuticals based on
classification system.
The
patents are classified by chemical structure and therapeutic intent, and a
company has mapped
not precisely one to one, about
which we drop.
map
pharmaceutical patents
use these data instead of the Hall-Jaffe-
Trachtenberg patent data because the latter use a classification
chemical structure, which
all
We
are
left
this
five
system into the
FDA
percent of the patents
system. Because the mapping
fall
with 275,406 patents. This number
is
into
two
of our categories,
comparable to the number
of pharmaceutical patents in the Hall- Jaffe-Trachtenberg patent dataset. Since
to obtain citations to construct a weighting system,
40
we
assign
all
we were unable
patents equal weight.
Table
Correlation s
Between
1:
Different
NAMCS
Panel A:
Drug Use Measures
overtime
1980/1990
1990/2000
1980/2000
Overall Correlation
0.955
0.878
0.852
Weighted Correlation
0.967
0.862
0.862
Mean
0.851
0.778
0.732
1996/1997
1997/1998
1996/1998
Overall Correlation
0.996
0.996
0.992
Weighted Correlation
0.999
0.999
0.997
Mean
0.919
0.955
0.902
Correlation by Drug
Panel
Correlation by Drug
Panel C: Correlation Between
B:
:
MEPScDvertime
NAMCS ami MEPS,
MEPS/NAMCS
and Between
MEPS Use
MEPS use/MEPS
Overall Correlation
0.187
0.962
Weighted Correlation
0.360
0.983
Mean
0.813
0.853
Correlation by Drug
Notes: Overall correlation
cells)
the
across
MEPS
average.
or
all
is
categories.
NAMCS. Mean
and Expenditure
expenditure
the correlation of use per person (or average expenditure share for expenditure
In
weighted correlations, observations are weighted by total use or expenditure from
computes correlations separately by drug, then takes the
correlation by drug
Table
Population by
2:
Age Group and Drug Approvals Over Time
Panel A: Log Population by Age Group, From
CPS
1970-1980
1980-1990
1990-2000
0-20
26.29
26.74
27.43
20-30
25.96
26.61
27.09
30-50
26.32
27.14
28.04
50-60
25.78
26.31
27.11
60+
25.82
26.58
27.30
Age Group
Panel
Age Group
B:
Log Drug Approvals of All Drugs bv Age Group, From
FDA
1970-1980
1980-1990
1990-2000
0-20
6.49
6.33
5.37
20-30
3.81
4.54
3.87
30-50
6.22
6.85
6.23
50-60
5.57
6.22
5.86
60+
6.22
7.21
6.38
Panel C: Log Drug Approvals of Generics bv Age Group, From
.
Age Group
FDA
1970-1980
1980-1990
1990-2000
0-20
6.32
5.90
3.95
20-30
2.64
3.74
2.89
30-50
6.06
6.68
5.68
50-60
5.33
5.88
5.36
60+
5.92
6.96
5.81
Panel D: Log Drug Approvals of Non-qenerics by Age Group, From
Age Group
FDA
1970-1980
1980-1990
1990-2000
0-20
4.62
5.29
5.09
20-30
3.43
3.95
3.40
30-50
4.34
4.96
5.37
50-60
4.03
4.98
4.91
60+
4.90
5.71
5.53
In Panel A, log income by age group is the log of mean income over the ten year interval, from the
March CPS. In panels B, C and D each of the 34 drug categories is assigned one age category, based
on the age group that uses that category most. See text and Appendix Table A1 for details. Log of drug
Notes:
approvals
is
the log of the
corresponding
sum
to the given
of
all
approvals
age category.
in
the indicated
1
year
interval, for
all
drug categories
Table 3A:
Effect of
Changes
in
(D
Market
(2)
Siize
on
(3)
New
Drug Approvals
(4)
(5)
Panel A: NJLLS for Poisson model, dependent variable
I
is
(6)
count of druq approvals
6.04
4.95
6.16
5.16
5.51
4.02
(1.95)
(1.90)
(2.04)
(1.84)
(1.84)
(1.51)
0.86
0.91
0.86
0.91
0.75
0.84
Log Market Size
R Squared
Panel B: OLS, dependent variable
is
loq druq approvals
4.52
4.65
5.74
5.27
3.41
4.39
(2.00)
(2.13)
(2.42)
(2.21)
(1.62)
(2.11)
R Squared
0.87
0.92
0.86
0.91
0.80
0.86
Number
204
102
204
102
204
102
5
10
5
10
5
10
Yes
Yes
Yes
Yes
No
No
Yes
Yes
No
No
Yes
Yes
Log Market Size
of Observations
Length of Time Interval
(Years)
Drug Category Weights
Market Size and Weights
Include Income
in parentheses. Counts of drug approvals are computed
Drug Approvals, by counting drug approvals for each category over five- and tenyear intervals (see Appendix for details). Market Size is obtained by multiplying the time-invariant average
expenditure share of users in a particular age group, calculated from the MEPS, by total income of that age group
at that date, from the CPS, and summing over all age groups. When market size does not include income, use
per person is multiplied by population. See text for details. All regressions include drug and period dummies,
and the 34 drug categories constructed from the MEPS, as described in the Appendix. In panel A, the Poisson
model is estimated by NLLS (with the Hausman, Hall and Griliches, 1984, transformation). See equation (26) in
the text. In panel B, if a cell is empty, log approval is set equal to zero, and a dummy variable, equal to 1 when
the cell is empty, is added to the regression; see equation (24). Regression weights are cell size for the category
Notes: Huber-White robust standard errors are reported
from the
FDA
from the
MEPS,
dataset of
New
either total expenditure or total use.
Table 3B:
Effect of
Changes
in
(D
Market Size on
New
(2)
Panel A: Poisson ML, dependent
Log Market Size
Drug Approvals
(3)
varisible
is
count of druq approvals
6.25
5.98
(0.52)
(0.57)
(0.67)
(0.74)
{1.53}
{1.50}
{1.81}
{1.73}
Number
of Observations
is
count of druq approvals
6.56
6.18
8.58
7.90
{1.74}
{1.91}
{1.66}
{1.70}
Panel C: Negative Binomial ML, dependent variable
Log Market Size
6.23
6.55
Panel B: Weiqhted Poisson ML, dependent variable
Log Market Size
(4)
is
count of druq approvals
5.40
5.11
5.39
5.27
(1.58)
(1.70)
(2.14)
(2.30)
{2.07}
{1.64}
{2.87}
{2.25}
204
102
204
102
5
10
5
10
Yes
Yes
No
No
Length of Time Interval
(Years)
Market Size and Weights
Include Income
Maximum Likelihood standard errors in parentheses, and Huber-White standard errors in curly brackets.
Drug approvals, market size variables, and regression weights are constructed as in Table 3A. All regressions
include drug and period dummies, and use the 34 MEPS-based drug categories. In panels A and B the
Poisson model is estimated using maximum likelihood. In panel C the negative binomial model is estimated
Notes:
using
maximum
likelihood.
O
cd co
CD t-
m
a>
oo
o
r~-
CD
00 Tf
CD
CD
00
CD
o
a)
n
a
cu
"S
3
a
^
co
CO
CD
00 hCD CM
o
CO
00
m
O CD
r-
<j-
CM CO
CD CM
*' in
^
in CM
CD
CD
d
00
CD
CD
s
-
-1 1
en
>
o
i_
Q
a
co
a
3
iZ
.S
CD
tf CD
CD 00
1*^ *'
r--
o
en
00
00
O
O
t- in
oo co
co
o
CD
CO
oo
00
CO
o
&
LU
c
-Q
-Q
CO
i_
ra
TO
CO
g
co
"O
** co
CD
CO
co co
c
0)
a
C
*
o
in CM
JO
cz
CD
CD hCM -<fr
TJ
c
cu
Q
CD
00
co
CD
O
cm'
o
DQ
o
CM
CD
C <U
O N
<0
O
.<2 *c O ro
OQ. 3
tU
U)
=*
co
=5
<
n£
8
to
J)
cu
>
<U
„S
CO
2, °Ola
C
O)
to
°|
O
m
hCD CD
IT) CM
"* co
™
73
s g
CO
CM
"S
ro
ro
cu
=
to
.S
Q_
cu
$ c
ro
»
CM tCO CD
cm f^T
<*> CD
<? CM
>*
oo
o
CM
t:
co
^ =
co
tu
o t
Qr
W
^
CO
CD TCM
CN
CD
"'
0)
CM
O
-~
N °co "a
CD i-
m
oo
"* CM
to
™ d
2
CD
TJ v-
J=
"D .E
~
i-
CM
C
CU
5
a *1™
c -a cu
2 <U
8 "R
=
TO
CO
o
~
g I
?o|c
O
3 °
—C
CD
CO
i
"a
aS
CD
>,
cu
ra
«? CO
m
o
CD CD
e
Z
* CQ
_
_ CO
—
x = c
CO
Q
ro
8
S
CL
CD
00
tj
TO
co
c
CD io
E cu
o o
«iJf'= 2
a>
cu
cz
CO
'"tj
£=
CO
CO
_l
CD
00
co
co
'-
cu
TJ
CO
ci
TJ
cu
<
^
>
"i_
a>
CD
^- in
CO 1CM
0)
_CU
>
c
"3-
O
CD
CD
CO
.2
m
'
en
•4—
.3
|"S 5
O
1^
§
c
3
O
o
cu
-H I c
™ S w §
<r
O
oo
T-
CD
CD
a
"5
O
«
_i
o
CZ
2
to
CO
r^ co
CO CM
TJ
"to
r5
™ co
C
„,-
a)
i—
rn
*-*
—
CO
CO
to
,u
.
t=
O
to
cU
S
CD
O
CM
"2
CO
"co
o)^S
j^
CU
-1-
O
rn
CO
c
o
fa
CU
N
cu
N
CU
b!
CO
CO
•*->
^_,
"cu
CU
CU
.*:
2*.
J£
i_
CO
CO
s
i_
CO
5
O)
o
2
O)
CO
"D
CO
CU
a
OL
N
cu
</>
CO
CO
CO
"cu
"cu
CU
J*
cu
t_
CO
13
cr
CO
cu
.a
a
cu
L_
CO
3
tj
CD
o
CTJ
CO
CO
cu
cu
£
cu
N
CO
CO
£
cr
CO
cr
O
cu
£
> ^O
™
S ^
-§"£
^£^
g
XJ
TJ
E
h-
cu
cu
-Q
D)
C
CO
CU
eu
g 9- E
o c o
—- ~
co
=!
o.EZ
g
E
in
m
CM iCO t-
CM
CD
CD
00
CD
00
O
-^l-
r-~
r~-
r~-
-*
00
oo
.9 cd
cn -o
o
o 3
o o
z
ffi-E
co
>
o
Q
Q
CO
a
3
4—
C
g
O
c
D
O
o
CD
—
m
o
co cn
en en
lO T-"
00
CO 00
oo
LO T-^
co
co
CD
*
>
o
c
o
o
CD
o~
en
*
T
LO CM
CD
cu
-I—
CD
CD
Q
g
CD
CO CO
CM CO
co' CO
en
CO
c7
T—
CO
CD
r-^
T—
_o
CO
P
00
CO en
oo
—
"S-
a>
>
h-
-~
-«—
C
r^
LO
CD
>
C
-•—
ac
CD
CD
CD
oo
CD
CM CO~
h- 00
LO
^^
"aj
"O
O
E
c
o
10
to
"?
-d-
CD
00
CD
co
r--;
LO
T
o
w
D
L_
O
c
3
O
o
o
CD CO
\f LO
CM
CD
CD LO
CO
*
en
oo
O
O
CO
CO
CO tLO
en
oo
o
o
CO
CO CO
CO
*
en
oo
oo
r--
co
oo
to
0)
c
o
CD
c
o
CD
a
CD
"O
L0*
c
CD
-<t
oo
ci
CO
CO
CD
L.
a
O
CD
c
o
X!
c
a>
D
00 GET
CO CM
*
en
oo
d
*
o
z
CD
c
o
ai
FT
CD
CD
"O
oo
en
0)
tf ^3,
'
>
C
-*—
CD
o
O
E
c
o
to
CO
LO
£
ai
o
c
o
z
0)
Z
o
'
i_
co
CD
a.
M-
o
"Z
'i—
CD sT
CO <—
CD
cu
CO
CO
CD
(D
co
oo
e
Q
Q
a
_CD
i
CO ccT
CM oq
CO ,-i
ac
"D
CD
X!
CD
"i_
CO
o o
e
o
-*—
a
*
CM
CD
°
M—
=3
i—
o
i
o
"O
i_
>^
Q.
Q.
a
Q
TO
a
CM CD
CO CO
1^
o
co
CO
"E
LO en
CM
•* CM
o
CO
CO cm
00
00
CD
H
>
a>
-I—
c
CD
c
CM
•<*
-Q
_c
**—
CO i00
CD
co
^_j
tO
m
o
iq cd
"L.
c
LO
oo
CD
co
CM oo
_l
OQ
CO cm"
I-- CM
CM
oo
oo
*
oo
oo
co
LO
o
z
<U
>-
CD
C
o
z
"53
ZJ
c
C/D
a.
co
00 CD
cm h-
LO
CO
o
o
oo
oo
CO
T—
o
oo
h- co
CD
oo
oo
c
o
'-4—
c
CD
-I—'
O
TT LO
<-:
cb
oo en
CO
LO T-^
CM LO
LO CO
CD CD
CM
en
CD
CO T-
CM cn
CM
CD CD
CO
CO
CD
co ST
cm en
t'
O
LO CM
co co
en co
t-^ CD
CO
en
CD
co
"*
co
^ 9
o
o
CD
o
c
o
CD
fj
*
C
o
CD
CD
a
£
C
CD
CD
N
00
CD
CD
a.
Q.
4—
CD
J*
O
N
E
CD
a
°
"tD
(D
-S
-Q
CJ)
fD
CD
>
-I
tD
CD
h_
CD
N
T3
C
CD
•a
Q
3
CD
-Q
cr
cn
2
co
>
W
cr
O
_i
cn
g
C
3
E
"CD
ro
i—
CD
i_
CD
CD
cr
T3
3
W
o
3
CD
CD
CD
"a
_3
cn
CD
CD
CD
"a
o
CD
c
D"
TD
i~
-^
D)
to
m
CD
CD
>
LL
Q.
CO
CD
hig
iS
CD
a.
LU
cn
s
t
to
CD
3
s
'i_
o
cn
CD
c?
a:
o
CD
O
cn
CD
O
X
-»—
^a 5
a>
CD
O
03
O
CO
a
CZ
3
i_
•:=
Q_
=3
co
03
O _t5
i=
cd
CO
CD
c
73
_3
-'
»5
cd
co
CD
>»
"
ac
co
CO
5
—
en
"°
CO
c 2
CO
73 >-
£<
o
en
03
To
O
o
- ^ o c
—
CD "8*73
U
co
o
CO
-Q
P -tS co
£ CO
c
CD
'i_
. co
3 O Q)
O d)
Q-n
CO
c £ L: to a>
o
CD
>,*"
0) O
c *" en
73 ^ ~
03 CT CO
c
•- o o
It o U)
co
i is 1 c DlO'oi
C 03 O
°
73
TO
O £
T3 03
Si
C
as ns en
o
O rn
3
,£
ro l<03 ==
CO
2 §
z
73
3f
oS
»T
'.
en
73
.
03
oo
o
o
03
L.
TO
CM „ro
CO
£
JD
XI
CM
2
co
O
> —
C
CD
b!
<o
o
o
•
co
^ n"
<"
u-r
1
co
E
73
CZ
co
jo
co
>
o
L.
c
CO
—
CD
12.
03
0)
T3
T3
03
X!
•
<"
CD
=
L.
CD
03
g°
CD
CO
•o
CL
<
^o
o
CD
S
CO
03
03
73
rn
b"
<
a
LL
CM
ro
O
03
73 73
<*
tr
° CD
0) CO
C
O
OS
C33
f-
c
E
_
en
°i
£ 5 £
c
« 2" E
p 73
Q.
ii E
° E *" c 2 "E 15
£ .2 §
>t = £
en c c E 73 T3
o 3
o c
^ £ s ™C
3 ~*
C 3 £<
73 -a CO £
H oen
o
•;' _Q
° a
73 3
C CL
o
E < "S
To
e"
8 « 2 '^ m
b5 c
-
sz
—o
s:
£
03
o
oo
CD
>,
CO
I—
CD
CZ
03
D3
en
in)
0)
03
CO
CO
CD
o
en
5
03
"co
o
en
cd
3
03
co
01
oo
CL
CL
5
.2
CO
-^,
T3
CL
(U
1—
03
03
tr
Q.
D3 Q.
CO
£ C
8 8 I
CL
CL
C o o
CD
«
« g>=
roiS
3
T3 JO
o .2,5
o 03 P
X
-S T3
t>
E
CD
L.
L_
co
en
CO
03
CO
03
£ 0,B
03
Si
£
851 S.W
CL
Q.
ra
co
o)
.2
c
3
03 !2 o
CD
E S "
E
-2 CO
03
o
«
5 £ "^
-2 "S<
c S oj
= Is
< 5 *rf
03
CO
E
S
>£=
co
c
oo
CN <o
T~
CD
-«—
o>
^
8 E
o
- .2
_^
O ^
'
JD
X3
03
CO
L_
v_
Jo
.cz
co
CD
,-,
03
CZ
CO
03
cj
CO
-_
.52
cu
N
CO
co
0!
CD
££
E
3
C C
>
3
I
03
.
-£
03
c
co
-
jD
-Q
CO
^
03
2.
CO
-Cl
03
i=
73
r-
£
O
Jj
"53
03
CL
CD
73
73
03
CZ
co
en
en
_
CO
—
=
cn £ ^ CM
£o —
2.2>73 73
_
£ c
CO
!(?
03
CD
"a
03
•
CO
03
z:
TO
5
o
03
e:
03
jo
CZ
03
Q.
JZ
TO
03
CO
=> 73
CD
to
O-
CO
>
en
.t;
0)
0)
CL
73
03
73
03
_
C
O
CL
CD
CO
CD
73 •^ -^
CD
-t;
Z-Z
o
t en
0)
o
CO
o To
03
-° CD o
_cz
.C .CZ o
o
O!
—
JZ
co
03
O
.2 co
^s
c:
£
CO
73 73
CZ
CD
i_
CD
co
0)
CO
-^
CO
o
CD
cn en
03
o
CO
L.
03
c
CO
CD
co
CD
73
c
en
M
iJ
CD
Ol O) Jl
i- t- J=
oj
73
C
03
CO
CD
-Q
co
"u.
£•73
03
TO
CD
CO
03
cd
l_
I—
CD
r
n IJ
o i:
73
CD
CU
l_
m
oo
*~
CM
CM
^f
CD
co
5
co
3
ro
o
o
CM
S £
CU
cu
CO
cu
cu
CO
ro
-
cu
3 u
X '5 t ro
cu
> o
D.
-4-J
-o
cu
co
ro
CO
en co
cm
CO
CO
CD
00
o
ro
g
O
CM
ro
o 2
>
o
i_
Q
Q
i
ro
CO
CO
CM CM
1^
co
CM
cu
>*
o
CM
E
O
C
<u
CO
LO
-<t
JD
co
in hCD h-;
co cm'
£
03
!_
CO
CO
co
O
in
g
T3
&
CU
E
£
in
CO
co co
CO eg
t-
O
V
m
CM
CO
CO
o
en
cm
co co
-^
ro
g
,-:
CD
to
cu
c
cu
a
o
E
c
o
h- cm
*
ro
CO <*
co en
CO CM
00
co
CD CM
co
co co
co co
co
00
d
>
o
CO
c
o
<
0-
O
*
p
CO
oq
CO CN
co
CO
|^.
(v,
V
co
O
3
to
c
ro
00
CO
CM
"=f
o
CM
g
cu
c
ro
c
o
g
0.
cm
CO CM
i-
en cm
r-- en
O
co
co
fi
CO
o
CM o
co en
en en
r-^
co
'ro
g
O
CM
T3
z£
iiXJ
-D
CU
CO
ro
£
ro
cu
Q_
1
^'io
cu
ro
-O
ro
h-
CU
c
5 E
3
<U
ro
o
CO
ro
o
tj
£5
ro
O
en co
co en
CO
co
CD CO
co co
CM CM
00 CM
LO CM
Noo
CD
>
^
o
CM
c
<
Q-
E
o
o
.9
ro
cu
CU
.N
CO
1
,
cu
\s
i_
ro
^
en
o
N
CO
cu
.*
L_
ro
5
a
ro
cu
CU
cu
o
c
CD
ro
CO
3
-*—
<o
c:
N
CO
cu
ro
cu
X
z.
ro
=j
cr
CO
a:
N
CO
•*->
ro
cu
cu
_*
ro
2
^
2
CO
O
u.
ro
T3
ro
cu
to
-
c
CO
a ^c
X 2
TO
CD
s
CU
I—
ro
cr
CO
cu
c:
ro
i_
3
_c
4—
: o
£co
c
t
ro
CU
CO
ro
3
O
™
o. jn
J2
ra
cu
o
co
<u
-ti
ro
ra
w
£
^
cu
cu
CO
i
b!
co
<"
^f CD
= «
a a.
ro
ri
ro.9
co
cu
CO •F
"O
c:
ra
cu
CO
3 -4=
o
o E
> cu.2^
° a
CO
Si
ro
cu
j:
4—
o
cu
~o
.!=
"O
"°
.
"g
g
ro
CO Tf
T
ro
*I£
to
ro
3
r-
=9
0)
CU
.9 "a
«-
•-
-
1—
i—
m
^
co
ro 4= (DID
2^ Cc £
to
E
o
c
ro
3 C
O ra
O Q. 3 «J
ro
E
to
m
a>
3 V
CU .9
ra
t e g cu ro
c o ro ro
co
o o. o 5 X
*-
_
.E
~
c
g
-
CO
cu
S
to
cu
cu
-a
^x
cu
cu
E
3 O 3 p co5
u co •a o~
CU
10
ra
^
"a
ro
a>
cu
9ro
co 9-
aE
ra
l.
cu
^ £r E
co
ac
Q)
CJ
>
o
o t
9-
.
ra
E
ro
m
ro
cu
co
ro
Q.
o
c
ro
a)
cu
-o
co
_
ro
O £
3 N
CO
-C
CO
£ —
c
cu
cu
o
c
ts|
I
cu
cu
cu
CO
CO
cu
cu
4-J
TCM CO
LO CM
cr
CU
™
u
ro
CL
_
X E
3 cu
o o 9
3 O
c
V—
cu
<"
'C
cu
CO
cu
-C
cu
CO t,
X § E
°.
<
CL
CO 8 2 3
to O
_cu
-a
CD
i S
CX CO
3 C
o g
^o ^ —
ro
co
cu
£
ra
cu
X
c
ra
E o
o g <=
Qj
u E
N
« 8 .9
CO
H
£ .g COO —
nj
en"
"cu
E
o3
"S
o
CL
_ «
ro
CO
CU
J3
-o
•^
So
co
CU
o
CU
E =
ro
o -i
Q.
cu
cbS
o
c cu
ro
t_
CO
o
o .g><
cmIV
O ^3
^ F
B-
ll
ir
ro
r--
>-
to
.rz
^
c
cu
cu
CO
cu
CO
oo
oo o
2
°? p
co
o
o
T3
C
~
-i=
"O
D)
.2
cu
CU
co
CD
o
ro
"cu
o
]>.
ro
c
cu
Q
CD
co
*i_
cu
X
.E
*=
•a
03
CD
cu
ro
>
c
CD
o
in
CO a,
ro
2 "O
a. cu C
o co
S o
Qcu
C
CD
jz
-4—
cu
<
c
o
o
CO
CO
c cb|
3
CU T3 H- CU
ro
i_
CO C O N
ro
ro cu
to
> w Q. CU
a. *ro x >. cu
ro
a
I
2-
>
CU
JO
™
ro
i—
cu
>
^E
cu
co'
§S
-o
»S^
-a
ro
CO
co
en CO
*
CO
CO
CO
co r~CD ID
O
,
o
«™s
co
cu
CO
<d
>
cu
»- -a
to
5
c o
>-
E-E
^:
en
^_-
CO
CD
5^
r--
o
<
CO
cu
o
T3
>-
<U
10
TO
XJ
CO
>
o
1_
O Tco m
^
o
co
co
<* cm
CD co
csi
o
o
o
o
en
co
o
Q
Q
CO
CJ
03
3
i_
O
"
o —g
«?
O
h-
.<2
CO
CD
C£
T3
CM
oo
o
<
I s
co
o
en
CO
CO
00
csi
co
o ^c
m
cd
en en
CO
co
en
o
<
o
o
en
co
(0
(U
CO 00
co
oo
CD
o
o
oo
CO
CO
Dl
LU
LO
o
2
2
TJ
c
>
CD
CD
00
cd en
r- co
CD
00
CO
o
o
o
co
CO
o
2
LO
<£•
CO
0)
J.
TO
cu
CD
>-
o
co
i
<d"
CO
CO
N
H
tt
II
O
to
c E
TO
.
CD
LU
CU
CD
.n
-•—
i
£
Eg
LO CO
en
o
D.
Si
<m
CO
£Z
CD
CD
CO
c C
o o
CO
to
CD
i—
cn
CD
<
CO
X a.
=H LU
5
*=
I
X
<=
<D
a o
T3
CD
^
3
SJ
3 ? S
co
cu
to
cu
>-
>-
o
>.
a>
co
CO
co
CD
_3 >>
3O o
i-
5
<° H.
Q. £
T3
O < E CD £
X
5 s c
± <
C T3
°° —
C
E S
3
O H
J=5
° £
a) «£
"| o
E
CO
C=
CO
CO
CD
o
ten
LO co
CD t^r csi
£!
™
CO
D.
LU
^-
TO
to
a)
>-
to
TO
c
CD
E
zz
L_
CO en
CO CD
LO V—
CD
00
roo
O
o
CM
co
CD
CO
CO
D.
UJ
co
to
cu
>
-a
Q)
-S
<
2
£
S."
*—c N £
CD
O
CD
-I—'
CO
rz
CD
cu
^r LO
en
o
CD
oo
o
o
CM
CO
rco
o
CM
co
—
co
c
o
•i=
0]
cu
cu
N
N
co
co
.a
CO
*-»
(D
-a
L_
CD
CD
i_
CO
cu
1_
CO
en
cr
CO
cr
CO
^
2
o
3
3
oc
en
o
oc
CO
Q.
LU
cu
co
I
E
O
M-
E
3
i_
o
.a
£D)
E
3
1
™
<u
S
"cD
CO
ro
TO
a>
^
CO
cu
O O
i_
CD
1
>=
£
(U
a
CO CO
™
£
CD
>-
C
CU
CO
~
n
CU
'50'
CO
Q)
co
2 <u
N
CO
O
CO
1
3
D-
-S
r-
co
IS
CD
"S
(D
C
<"
05
i
i-
c—
CD
±=
N
g
CO
a
S
CD"
§1
co
—
co
CU
N
S=
^:
.-ti
CO
5
v
a)
I »
E
a E3
CD
CO
i:
(0
s n
T
o
^
CO
en ^
en
.2
II
'<"
S
<D
co
CU
co
ro
£?
CD
CD
DC >.
.
1)
"co
CD
•*-
TO
.£
E
P
TOTO o
CD
o
CO T_
O
< „.
-si
CD
j^
CD
CO
i-
3
-C
TO
co
"o
£
-g
-t±
CU
CD
TO
3 &
TO
Q--a
to
cn
E £
"^
£
o c
-Q E -2 co O CD CD
3 3 to
CD
I o ora Ec
° 3 3
E
CO
cn co
a
5J £
-2 £
o
c^.E 8 ,«2
cd
CD
c
o
CO
°
^ UJ £
<"
gLO C° >!™
^
CD
<u
to
a>
£
is
E en
— .S3
i JJ
£5 § S
uj
w
2
II
1?
N N
CU
_>*:
co
CO
E
ro
co
CO
to
cu
>-
TO
a>
co
in
cn en
>>
11*
II
o o o
o
CD
CO
D.
HI
>d-
CO
TO
TO
CO
CO
CD >*
en en
E
CD
T3
CD
"S3
TO
LO
CO
CO
CD
O
X
a>
?CM n
CU
CU
c=
o
.2
DC
CO
>-
<
CO
^r
CM
S
CO
g.«
"cd
c
CO
-^J
CU
N
CD
—c
CO
O
2
<
N
CD
CO
3
CL
(U
a)
CO
|
o •o
3
o
CO
c
CD
-t—
CO
Q
CO
£
E
o
o
to
cu
a)
TJ
.E
CD
a>
co en
i- LO
cu
"D
C
to
C
c
H
CO
CO
>-
CD
>
c
c5
c E
TO §
.a
ID
co
1CD
CO
i—
—
2o
_co
'to
I—
T
00
to
"c
3
o
o
CO
Ol
TO
i i
E
O to E «
O c
E E E ^=
CO 3
CO
en
oq
^
«
")
_
o E
CO x- Z
to cu
c
CU
^ to
D) => b
3
CU -^;
"
DC
.2
CO
O o
o
o
en
o
cm
co
CD
]-.
cu
c
3- oo
CD LO
Tf CO
CU
i
-
-8
H
O
<
2
o
|S
"O
CO
O
O)
CD CO
cni ,u? LO
LO
2
JO
03
>
O
i_
Q.
<
r-
LO CN
T— CD
LO T""
CO
«
CT>
r^.
S"
oo
^j-
o
CN
CO
CO
o
°
CO CN
LO
oo
o
co-
c
=3
o
o
CO
'i—
co
>
c
>
c
-t—
O
t-
00 LO
CD
LO
*
*
LO
00
-I-
CO
*' CD
O
c
CO
a
c
"O
00 CN
t-
CD
CO
O
CO
_J
o
m
CO
c
co
cu
cu
D-
73
-Si
TJ
ZJ
CD
CU
o
C
"D
O
Tcd co
o
CN
"53
c=
t=
co
a
<L>
cu
a.
co
"C -S
3 "2
T5 .2
g.E «
Q_
a.
co
CN
CO
£
CO
i;
5>
en
00 LO
C
3« -c
0.
LO ->JLO CO
oo io
CD
-|S
»|
CD
oo
o
co
o
CO
o o
a
CD
co
&
LU
co
c
.2
>
-p
g
r
<0
in
CU
J)
D) D.
II)
r
o
a>
CU
<u
CO
<U
"i_
c 2
CO
CD
CO ™
Ic3|
co
"O
o
eu
>
£
.g>"£
2 E
a. c co
a. t o
TO
ss
cd
CO
"O 'C
•=
(0
CO
0)
*_ "O T>
° a "a
a
O
CO
ro
cd
o
CN
CD
CO
CO
ro
CD
-a
cu
CO
CO
0)
2
T3
_CD
CO
™ °
XI
CD
CO
CU
CO
rr
00
*j
o £
CN
a)
N
i
™ .-2
1_
£ o
o K
CO
O "a
c
2 CO
«
N £l?
Et'l
0)
O —
c
pr
CO
E2 cb
o
*
oo
o
>
2
Q
Q
nj
o
z>
<D
oo
jo
3
5
oo
00
CO
2
<
CD
.
ni
D
"S
CO
°-< 21
c co n
o
r-~
h-'
CO
oo
en
cd
CN
CO CN
O *
00
CO CN
o
CN
5^
™
CD
CD
V)
c
g
c
N
X3
CO
CO
CO
*-J
CO
-!*:
L_
CD
^
CO
o
_l
N
c
CO
CO
Q
"a JD
(0 X!
CO nj
en 'i—
en
CD
>
i_
CO
2
a
CD
CO
_i
"co
.*:
a
CD
CO
i_
CD
CO
cr
CO
i
2
•a
co
CZ
CO
a>
3
CD
_l
CO
N
CO
CL
_i
N
a.
O
-*—
N
XI
o
CO
CO
CO
a>
Q
CO
CO
5
T3
CD
CD
CD
CO
>
T3
CO
I—
CD
ZJ
X!
—I
CO
V)
CO
CO
Q.
CD
CO
CO
Ǥ
«
c g -a
CD
P O
CO
CO
> 2
»
<
£
CO
CO
CO
</)
u_
CD
CU
CD
c
CO
CO
«
2^
co
CD
cr
CO
«
XI
ZJ
c
CO
XI
CO
CD
CO
i"
CO
f!
"2
CU
in
CD
j
CO
co
-^
xi
,CD
Zj3h
1-2
c
CD
o
oo
CO t-
o
c
CM CD
£
CO
CO
CD
S
"CD
53
co
0_
"S
CD
CD
X
s^
Q-
a
"O
en .Q
CD
CD
CL
H5
CD
s-
CO
CO
CM CM
t- OO
CO cm'
C
s-
I
I
CM
-5J-
cn
o
CM
£
to
CO
>
o
t- CO
c:
_c
<r cm
o
CO
o
o3
c
s
I
"S
CD
>
o
-:
9
Q
&_
o
_co
Q)
o
CM
co
6 5
CM
CO
CO
CD CO
•*
CO
O
^
CO
cd
e:
Z
0-
co co
o *
E5
1^
tf
Z
u>
to
CD
in
co
°
m
CD
co CM
CM
a>
00
00
d
-s
CD
<
'i—
CD
Q.
E
T3
T3
a)
CD
^
en
en
co
CO
CO
CD
3
o
co
>.
in
CD
~1
>cj
CD
o E
CD
"O 5
CD
c
CD
I?
in
CO
"CJ
O
CO
o
CD
CL
CD
co
-
'i_
E
I—
c o
C
CD
c
CD
•D
C
CD
a
N-
o
CO
a>
CO
CO CO
CJ)
m
o
T-
CO
CO
>?
o
"O
£
LU
to
_i
_i
z
<
co
o
c >o
z g
i
"O
c
CD
D
CD
CO
co in
CD
o
CD
oo
CD
o
CM
< O
CD
CO
c
CD
a.
(0
>
c
-I—'
o
CM
CD
D)
CD
CO
o
c '^
o
Z g
i
—
I
CD
"O
i_
o
_g
co
|<
•3-
o
CM
^y3 ^,
CD
to
c 5
s b en
CO c
O CD
>
O
cd
e:
D-
t- CO
CO CM
CM CM
cd
CD
"
co
c
CO
cd
-o
jco
"cd
(0
o
CO
N
-i—
"I
en
cd
c
o
O
§_
£
g.E a
g
CD
CO
"a
<o
o
c 'c:
o
z g
i
s
in
CO
co CD
CO
*'
f
^
o
CO
00
co
cm co
CM
<
"53
c
CO
co
co
CO
,
—
O
CD
J2
CD
co
ti
™
>
-tC
s>
.51
co
CD
cd
-a
"D
,„
CD
CL
Si
-a
CD
CD
CD
"3
CL
CO
m
m
T
CO
-* T^
m
co
T~"
T—
d
o
m o
o
1
CO
o
CO n,
CO CD
CD
oo
9 a
d
z C
< O
CD
O
CO
CD
QJ
CD
-C
C
CO
o
<u
r;
„
£
CO
CO
a,
CD
£
CD
(0
c
co
CO
CD
r-~
co
co
i-~-
^r
00
d
o
c 'i_
o CD
Z C
Cj
o
CM
CD
< o
C
E
CO
c
o
CD
a5
x:
l_
CD
c
2
CD
-a
-4—*
CD
N
CO
"55
i_
CD
^
o
_l
"co
XI
C
CO
CD
CD
CD
_l
•a
CO
"O
a.
CD
Q
-a
CD
-Q
CO
CD
"53
g
CD
en
en
_i
N
CD
'i_
CD
3
0_
CD
CD
cr
CO
> O
CO
VC
CD
XL
i__
CD
C N
CD
">
<^
en
o
_i
X5
CO
c
CD
CD
*t_
CD
_l
"a
TO
CL
CD
a
a 5
XI
"O
o
*i_
CD
CD
-i
c
CD
i—
CD
"D
cr
CD
CD
c
C N
CO
Q.
CO
tx
cd
';=
CD
CD
CD
O
>
-t—
3
CD
en
en
CD
CO
CO
>
CD
Q
<->
s -s
O 3
t O
CD
e
c:
CO
"S3
&
a
CB
n
w
£
CD <_,
CO
c
cn
Q-
CO
03
<U
CO
^
E
? o w
u ^=
9?
cd
CD
E XI
O b
§8
o
CD
0_
CD
CO
$&
"*-
<
Q Q
03
E
O
CO
</)
£
XI
CD
CO
"a |CO
—
tu
N
!S
ro
«5
CD"
^1
o
CO
CD
£0
<-)
c
>
t;
CO ^j
XI T3
—ci
"S §
CL
CD .£
"D x: cd
en td
XI
CD T3
CD
_CD
g CD
.a
> TJ
f CCD co iS
§ C
CL T3 CD
CD
'D TD Q. c o
C < ro
'J
s
I
s-
-,
^
cd
ria
<U
CD
-a ^2.
E g
E T3
3 CD
"a c
« £
CD
£ >>
E
co
g o
co xi
en 3
c i_ CD
CD CD
3
co
CO
O
z CD ocj CD>. <9o .i=c
o n
I
CD
CD
E
c=
'
>
,
Appendix Table A1:
Summary
of Classifications
and Drug Use by Age Group, Computed From
MEPS, 1996-1998
Use Per Person
(
Share of Use
in
Parentheses )
Aae Group
0-20
Class
Description
10
Lincosamides, Sulfonamides,
Penicillins,
Cephalosporins,
20-30
30-50
50-60
60+
0.61
0.30
0.38
0.44
0.45
(0.40)
(0.09)
(0.26)
(0.10)
(0.16)
0-20
Misc. Antibacterials
Tetracyclines, Urinary Tract
0.02
0.06
0.06
0.09
0.12
Antiseptics, Quinolones
(0.11)
(0.13)
(0.31)
(0.14)
(0.30)
11
12
20
30
40
41
42
43
50
60
Antifungals, Antivirals
Hematologics
Cardiovascular
-
Renal
0.03
0.03
0.08
0.09
0.07
(0.16)
(0.08)
(0.43)
(0.15)
(0.18)
0.00
0.00
0.03
0.11
0.43
(0.01)
(0.01)
(0.11)
(0.12)
(0.75)
0.05
0.10
0.69
2.68
6.05
(0.01)
(0.01)
(0.15)
(0.19)
(0.65)
Sedatives/Hypnotics,
0.01
0.04
0.16
0.27
0.41
Antianxiety
(0.02)
(0.04)
(0.23)
(0.38)
(0.33)
Antipsychotics/Antimania,
0.08
0.19
0.57
0.70
0.57
Antidepressants
(0.06)
(0.07)
(0.46)
(0.18)
(0.23)
Anorexiants
Misc. Central
Nervous System
Gastrointestinals
0.04
0.01
0.03
0.02
0.01
(0.46)
(0.07)
(0.35)
(0.09)
(0.04)
0.11
0.01
0.01
0.01
0.02
(0.75)
(0.04)
(0.09)
(0.03)
(0.08)
0.03
0.08
0.24
0.52
0.83
(0.03)
(0.04)
(0.27)
(0.19)
(0.47)
Hyperlipidemia, Electrolyte
0.01
0.01
0.13
0.67
1.37
Replenishment/Regulation,
(0.01)
(0.00)
(0.12)
(0.21)
(0.66)
30-50
30-50
60+
60+
50-60
30-50
0-20
0-20
60+
60+
Calcium Metabolism
61
70
71
72
73
80
Vitamins/Minerals
Adrenal Corticosteroids
0.06
0.09
0.07
0.09
0.13
(0.20)
(0.16)
(0.27)
(0.12)
(0.25)
0.05
0.04
0.09
0.14
0.24
(0.13)
(0.06)
(0.29)
(0.14)
(0.38)
Androgens/Anabolic Steroids,
0.02
0.38
0.34
1.30
0.67
Estrogens/Progestins
(0.02)
(0.13)
(0.26)
(0.33)
(0.26)
Blood Glucose, Thyroid
Contraceptives
Immunologics
0.02
0.10
0.35
1.02
1.70
(0.01)
(0.03)
(0.22)
(0.21)
(0.54)
0.01
0.21
0.10
0.01
0.01
(0.06)
(0.47)
(0.42)
(0.02)
(0.03)
0.00
0.01
0.02
0.02
0.03
(0.08)
(0.09)
(0.33)
(0.14)
(0.36)
with
Peak Share
Use
60+
60+
50-60
60+
20-30
60+
of
Appedix Table A1
(cont.)
Use Per Person
(Share of Use in Parentheses)
0.09
0.09
0.12
0.17
(0.26)
(0.27)
(0.12)
(0.11)
0.01
0.01
0.01
0.02
0.04
(0.18)
(0.07)
(0.27)
(0.11)
(0.37)
Topical Steroids
Extrapyramidal
0.00
0.01
0.03
0.03
0.08
(0.03)
(0.03)
(0.34)
(0.11)
(0.49)
Movement
Anticonvulsants
110
0.05
0.11
0.26
0.29
0.27
(0.08)
(0.08)
(0.44)
(0.16)
(0.23)
140
0.00
0.01
0.05
0.16
0.17
(0.02)
(0.03)
(0.25)
(0.27)
(0.44)
20-30
60+
fin+*
\j\j
30-50
60+
Oncolytics
Misc. Ophthalmics,
0.00
0.00
0.01
0.06
0.41
(0.00)
(0.01)
(0.05)
(0.08)
(0.86)
Glaucoma
0.06
0.05
0.05
0.08
0.16
(0.23)
(0.09)
(0.22)
(0.11)
(0.36)
0.02
0.01
0.01
0.03
0.04
(0.30)
(0.07)
(0.22)
(0.14)
(0.27)
Ocular Anti-Infective
130
131
60+
0.09
Skeletal Muscle Hyperactivity,
21
50-60
(0.25)
101
1
30-50
Infectives
91
120
20-30
Dermatologies, Topical Anti-
90
100
0-20
Description
Class
Aae Group with
Peak Share of
Use
Topical Otics
0.02
0.01
0.03
0.05
0.13
(0.12)
(0.05)
(0.24)
(0.13)
(0.47)
Vertigo/Motion Sickness
General Analgesics, Narcotic
0.09
0.29
0.59
0.89
1.13
Analgesics, Antiarthritics,
(0.05)
(0.08)
(0.35)
(0.17)
(0.35)
60+
fin+
T
UU
0-20
RD T
+
UU
30-50
NSAID
141
142
143
0.01
0.02
0.04
0.07
(0.03)
(0.32)
(0.17)
(0.43)
0.00
0.00
0.01
0.06
0.14
(0.00)
(0.00)
(0.13)
(0.19)
(0.68)
Antigout
Central Pain
150
160
0.00
(0.06)
Non-Narcotic Analgesics
0.01
0.01
0.02
0.02
0.01
(0.15)
(0.10)
(0.45)
(0.15)
(0.15)
Syndromes
0.00
0.01
0.03
0.03
0.05
(0.07)
(0.07)
(0.37)
(0.12)
(0.37)
Antiparasitics
Antiasthmatics, Nasal
0.20
0.14
0.22
0.47
0.66
Decongestants
(0.20)
(0.07)
(0.23)
(0.16)
(0.35)
Antitussives, Antihistamines,
0.15
0.20
0.29
0.45
0.41
Corticosteroids
(0.17)
(0.10)
(0.33)
(0.17)
(0.24)
161
Notes: Construction of the 34 categories
FDA
0.07
0.05
0.07
0.08
0.08
(0.29)
(0.09)
(0.31)
(0.12)
(0.18)
Cold Remedies
162
sub-categories.
Share of use
is
Use per person
is
is
described
the
the fraction of drugs used
in
in
fif)+
*
VJU
fin+
\j\j~
30-50
30-50
60+
30-50
30-50
the Data Appendix. Each category includes the indicated
mean number
of drugs in the class
the category by the age group.
used per person
in
the age group.
N
•
1—
in
on
03
<D
O
D
td
•
Oh
i—
o
H
s—
<
GO
O
&o
<D
J-t
E
o o
o o
to m (D
+
o o o
^
O o
CM CO LO CD
{tw*
CD
cd
CD
CD
CD
CD
CO
CD
CO
CD
CO
>-
CD
CD
CD
CD
CO
CD
CO
CD
lo
<*•
o
^r
o
LO
co
o
CO
o
in
CNJ
CO
C\J
o
T
o
ajeqs
o
LO
o
Figure
3:
Share of Income by Age Group, 1970-2000, from
CPS
—--0-20
-20-30
—A-
-
30-50
—X--50-60
-*--60+
1970
1975
1980
1985
1990
1995
2000
Year
Figure
Share of Drug Approvals by Age Group, 1970-2000,
from FDA
4:
0.45
0.4
0.35
—--0-20
0.3
—m- -20-30
£ 0.25
—A--30-50
—K- 50-60
—*--60+
re
w
0.2
-
0.15
0.1
"x 5
"
-"Sf
0.05
1970
1975
1980
1985
Year
1990
1995
2000
Figure
5:
Approvals Residuals vs Market Size Residuals
(J)
m
61
>
T3
CM
CD
ro-rZ3
;g
en
Q)
43
^ 73130
>0
jW
41
]$2Qi o
CO
>
13042
1—
Q.
Q.
\
'
61
13 °
4S>
<
p>,
CM -
61
>
Y6
43
Wffib
42
73
n
r
.1
Log Approvals Residuals
Log Approvals Residuals
Fitted
values
Notes: Log approvals residuals and log market size residuals are residuals from OLS regressions of log
approvals and log income-based market size on category and time dummies, weighted by expenditure
with five year intervals. Fitted values are predicted log approvals residuals obtained from OLS
regression in Table 3A Panel B, column 1
STata"
3337 312
MIT LIBRARIES
3 9080 02617 7516
',''
''. In
!
Iilnflil
'fill
:.
i
11
i
1