A discussion of hybrid open source and proprietary licensing models.

["Marshall W. Van Alstyne" <marshall AT MIT.EDU>]

On Sept 30, a colleague and I finally completed an economic analysis of 
optimal licensing that compares free/open licenses to proprietary 
licenses.  The key yardstick is the level of benefit or "welfare" provided 
to both consumers and to developers, while accounting for innovation over 
time.  We also analyze the incentives of profit motivated firms to participate.

This note summarizes key findings and is a request for comment. Pls feel 
free to redistribute, blog, challenge, etc, We just welcome feedback.

We plan to move the modeling framework into broadly accessible free / open 
code both so that the model can handle additional complexity and so that 
anyone can examine the essential tradeoffs (or modify them).  Our hope is 
to move discussion of optimal licenses into "open science" using a formal 
model rather than just formal argument.

The main question we address is:

    If proprietary incentives matter for innovation, and if public access 
to a reusable free/open platform also matters for welfare, then what scope 
of protection creates the greatest good?

KEY PARAMETERS
(i) Scope of protection is measured in terms of free access or "sharing" 
sigma, in the range of 0-1.
(ii) Scope is also measured in terms of duration or "time" for proprietary 
period t, in the range of 0 - infinity.

Before time "t" profit motivated developers can sell derivative 
works.  After "t" it becomes free / open by OSI standards.  Additional 
parameters include:

(iii) network effects
(iv) size of consumer base
(v) size of developer base
(vi) level of reuse
(vii) developer motives (ranging from "freedom motivated" to "profit 
motivated")
(viii) prices
(ix) intrinsic value of a platform or code base.

KEY FINDINGS

1) HIGHEST SOCIAL WELFARE

    a) The highest social welfare of any configuration is a pool of 
"freedom motivated" developers matched with completely free/open 
licenses.  This is higher than a pool of "profit motivated" developers 
matched with proprietary licenses.
    b) BUT, if any fraction of the population is profit motivated, then 
the socially optimal license will ALWAYS contain the offer of a proprietary 
period t > 0. This stimulates higher innovation.
    c) Regardless of whether an institution is freedom or profit 
motivated, the optimal license is more open than BSD in the sense that the 
length of proprietary protection on derivative works should not be 
arbitrarily long.  Multi-period innovation is thwarted by overlong protection.

2) PROFIT MOTIVATED FIRMS

The analytic framework allows us to consider stylized business models 
ranging from completely closed and proprietary (e.g. restrictive 
End-User-License-Agreements), through partially open access (e.g. allowing 
plug-ins and APIs), to subsidizing users and taking all profits on 
developer royalties (e.g. computer games), to non-profits that open all 
aspects of a platform.  We find that:

    a) Profit motivated firms rationally choose EULAs more often than is 
socially optimal.
    b) Decentralized innovation can increase profits over going it alone. 
For profit motivated firms, this means it becomes privately rational to 
choose open licenses once reuse and network effects pass a critical 
threshold.  This effect can even dominate subcontracts with targeted 
developers.
    c) Even non-profits can benefit by harnessing the efforts of 
profit-motivated developers and giving them a brief period to charge for 
the value of their investments.

3) PLATFORM OWNER TENURE

One open question has been whether large projects need "sponsors."  In 
general, we find that they do.

More specifically, we find that a platform sponsor needs to exercise a long 
term interest in a platform as a means to enforcing good behavior on the 
part of decentralized profit motivated developers.  A prisoner's dilemma 
emerges in which anyone who likes to charge prefers to do so as long as 
possible but, if they do, then the whole community suffers.  If a license 
can enforce a reasonable, i.e. short term, proprietary period then 
everyone, including the developer, wins over multiple periods of 
innovation.  For this to work, the tenure of the platform sponsor must 
survive multiple periods of tenure for downstream developers.

4) PIRACY

Even if no reuse is possible and innovation never occurs, conditions exist 
where a profit motivated firm prefers to allow limited piracy, interpreted 
as free user access, to a portion of its products.  Roughly interpreted, 
word-of-mouth effects on sales can dominate lost sales.


Three Points of Context:

1) Stallman has seen and commented on earlier drafts.  One key observation 
was that only a fully open license (where sharing parameter sigma = 1) 
could qualify as a "free" license and not anything that kept any piece 
proprietary (where sharing parameter 0 <= sigma < 1).  Critically, he's OK 
with charging 3rd parties in some cases: if a 3rd party takes free code, 
enhances it, and charges for their enhancement, then it's also OK to charge 
that 3rd party a modest fee or royalty.

2) This version is written for an academic audience.  So, it can be 
mathematically challenging, but if you want a "proof" of certain free / 
open source arguments, here it is.

3) The paper is available at: 
<http://ssrn.com/abstract=639165>http://ssrn.com/abstract=639165