A Model of Optimal
Software Patent Policy Design
Professor Matt Thatcher
Trade Secret Laws

Uniform Trade Secrets Act (UTSA)



gives right to companies to keep certain information secret
(to maintain competitive edge)
covers formulas, patterns, programs, devices, methods,
processes
Must have the following characteristics:





be novel
represent economic benefit to firm
involved some cost and effort to develop
is generally unknown to the public
company must show effort to keep the information secret
2
Trade Secret Laws (cont.)

Problems:
Software often must be put into the public realm,
making it difficult to keep secret (and generally
unknown to the public)
 does not protect from independent discovery


Economic Espionage Act (1996)
penalties: up to $10 million and 15 years prison for
theft of trade secrets
 IP lost in industrial espionage  > $300 bill / year

3
Trade Secret Laws (cont.)

How do you show you are keeping information secret?






identify all information to be protected
label it confidential
educate employees of importance of trade secrets
make only accessible to limited # of people on a need-toknow basis
develop non-disclosure agreements
develop non-compete clauses


Compuserv v. IBM (2005)
technology protections

firewalls, encryption, secure databases, etc.
4
Why Do Firms Patent Software?
Number of Software Patents Issued Per Year
[Bessen and Hunt (2004)]
Number of Patents Issued / year
25000
20000
15000
10000
5000
0
1982
Year
2002
5
Improvement to the Basic Product
Patent Width
(w = h - )
Patent Height (h)
Allowable
Imitation ()
Basic Product (s = 0)
Patent Height:
Patent Width:
Patent Length:
Protection from improvements
Protection from imitation
Duration of protection
6
Patent Laws

Requires that innovation under review must be:


new, useful, and non-obvious to a person of
ordinary skill in the relevant field
Once awarded, a patent provides patent-holder:
scope of protection from imitation for 20 years
 protects functions/behaviors of the program
 protects from independent discovery


Cannot patent

abstract ideas, laws of nature, scientific principles
7
Patent Laws (cont.)

U.S. Patent and Trademark Office (USPTO)

patent examiners (~3000) search for prior art
existing body of knowledge that is available to a person
of ordinary skill in the art
 what is the problem with this in the software arena?

determine originality and novelty
 ~ 25 month application process
 details about patented innovations are placed in the
public domain


No limit to monetary penalty
8
Debating Patent Policy Design

Why reform the U.S. patent system?


What is wrong with software patent quality?




Samuelson (2004)
patent height is too low!
patent width is too wide!
patent length is too long!
Economic perspective


National Academies(2004)
Federal Trade Commission(2003)
9
Research Questions


What is the target of feasible patent policy designs?
Which policy designs in the target




Which policy design is socially optimal?
How do changes to an established policy design affect
social welfare? That is, what happens if the authority




are good/bad for society?
are good/bad for consumers?
increases height?
narrows width?
shortens length?
Should software products be patentable at all?
10
Model of Duopoly Competition
(R&D Race / Product Improvements / Price)

Pre-game


posit a basic, well-known software product
 where quality of basic product is normalized to zero
Free Market Competition (No Patent Model)




Stage 1: two firms compete in R&D to develop a novel idea
Stage 2: the innovator (n) transforms a novel idea into an
improvement (sn > 0) to the basic product at substantial fixed
cost, C(sn)=ksn2 /2
Stage 3: the imitator (m) observes the innovator’s product
improvement and decides to what extent it will imitate at zero
fixed cost, sm  [0,sn]
Stage 4: the firms simultaneously set prices and offer products,
where MC(sn) = MC(sm) = 0
11
Vertically Differentiated Demands
(Stage 4)

Individual’s utility function:
U i  i s  p


where  ~ 0,1
Individuals’ purchase decisions:
 i sn  pn   i sm  pm
pn  pm
 i 
sn  sm
 i s m  pm  0
pm
 i 
sm
Solving for product demands:
pn  pm
Qn  1 
sn  sm
and
pn  pm pm
Qm 

sn  sm sm
12
Graphical Representation of Demand
(Stage 4)
sn
sm
pn
pm
 1
pn  pm
sn  sm
pm
sm
 0
13
Graphical Representation of Demand
(Stage 4)
sn
sm
CS
DWL
pn
pm
RRnn
CS
Rm
 1
pn  pm
sn  sm
DWL
pn
sn
 0
14
Solving for Prices
(Stage 4)

Firms’ profit functions:
 n  pnQn  Cn and

 m  pmQm
Profit functions (given quality levels) are concave in prices:
 2 n
 2 m
2 sn
2



0
and


0
2
2
pn
sn  sm
pm
sm sn  sm 

First Order Conditions (F.O.C.s):
 n sn  sm  2 pn  pm
 m 2 sn pm  sm pn

 0 and

0
pn
sn  sm
pm
s m s n  s m 

Solving the F.O.C.s gives:
pn ( sn , sm ) 
2 sn ( sn  sm )
s (s  s )
and pm ( sn , sm )  m n m
4 sn  sm
4 sn  sm
Qn ( sn , sm ) 
2 sn
sn
and Qm ( sn , sm ) 
4 sn  sm
4 sn  sm
15
Performance Measures
(Stage 4)

Firms’ profit functions (after substitution):

2
2


4
s
s

s
ks
s s s  s 
 n sn , sm   n n m2  n and  m sn , sm   n m n m2
2
4sn  sm 
4sn  sm 
Consumer surplus function:

2
(
4
s

5
s
)
s
n
m
n
  s n , s m  
2(4 sn  sm ) 2
Social welfare function:
  sn , sm    n sn , sm    m sn , sm     sn , sm 


sn 12sn2  sn sm  2 sm2 ksn2


2
2
24sn  sm 
16
Solving for Imitator Quality
(Stage 3)


Imitator profit function:
sn sm sn  sm 

 m sn , sm  
4sn  sm 2
Profit function is concave in quality:
 2 m
2 sn2 (7 sm  8sn )

0
2
4
sm
( 4 sn  sm )

F.O.C.:
 m sn2 (4 sn  7 sm )

0
3
sm
( 4 sn  sm )

Solving the F.O.C. gives the imitator’s best response function:
4 sn
s sn  
7

m
17
Solving for Innovator Quality
(Stage 2)

Innovator’s profit function is concave in quality:
 2 n
s

2
n
F.O.C.:
 n
sn

 k  0

7
 ksn  0
48
Solving the F.O.C. gives the free market outcome (MO) values:
7
1
7
1
,
) and ( pn , pm )  (
,
)
48k 12k
192k 96k
49
14






 n ( sn , sm ) 
and  m ( sn , sm ) 
4608k
4608k
196
259






 ( sn , sm ) 
and  ( sn , sm ) 
4608k
4608k
( sn , sm )  (
18
The Patent Model

Patent policy is set before the game begins:




patent height (h)
patent imitation level ()
patent length (t)
where 0 <  < h and t  [0,1]
Which policies give the innovator profit incentive to
seek a patent?
 nPO h,  , t    n sn , sm 


where  nPO h,  , t   t n h,    1  t  n h, sm h 
19
Conclusions


The POR is the target of feasible patent policy designs
Impact of changes to established policy design on social welfare




Optimal patent policy design




 patent length   social welfare
 but contracts the POR and eliminates good policy options
 patent width   social welfare
 patent height   social welfare if patent is long and high
max length and set height/width to intermediate levels
if length is short  set the highest, widest policy in the POR
reduces R&D incentives to discover a novel idea
Is policy too low, too wide, and too long?


it depends!!!!
many factors may prevent the policymaker from identifying the POR or
hitting a good
20
Questions?
21