Adware, Shareware, and Consumer Privacy by Nataly Gantman, Tel Aviv University Yossi Spiegel, Tel Av

1
Adware, Shareware, and Consumer Privacyby
Nataly Gantman, Tel Aviv University Yossi
Spiegel, Tel Aviv University

• NET institute conference
• April 1, 2005

2
Introduction

• Online distribution of software
• Shareware - software must be paid for once a
  trial period expires
• Adware - ad supported software
• Spyware - software installed without the end-user
  consent and tracks and collects personal
  information without consent

3
The main features of our paper

• We study the choice of programmers between
  shareware and adware
• Strategic interaction between programmers, firms
  that may advertise their products through ad
  banners, and consumers who buy software and
  products.
• A main premise of our analysis is that targeted
  ads sent to adware users is not only a nuisance,
  but may also be beneficial
• Claria Corporation has over 900 advertisers
  (including 85 of the Fortune 1,000), and an
  annual revenue of 90.5 million in 2003 and a
  profit of 35 million.

4
Related literature on the economics of privacy

• Posner (AER, 1981) - Privacy can be
• Concealment of information
• Peace and quite
• Freedom and Authonomy
• Privacy as "concealment of information Info.
  on consumers preferences allows firms to
• Use personalized prices (Acquisti and Varian,
  2004 Calzolari and Pavan, 2004 Dodds, 2003
  Taylor, 2002 and 2004 and Wathieu 2002)
• Offer products that better meet consumers' needs
  (Varian, 1996)

5
Related literature on the economics of privacy

• Privacy as "peace and quite"
• Hann et al. (2003) - firms invest to identify
  potential consumers, while some consumers
  (privacy guardians) invest to avoid
  solicitations. Competition between firms raises
  both types of investments (which are socially
  excessive) and hence raises the cost of privacy
  protection.
• McAndrews and Morgan (2003) - phone users can buy
  caller ID service to block telemarketers, while
  telemarketers buy ID blocking
• Hann et al. (2002) - find empirical evidence
  that, among U.S. consumers, protection against
  misuse of private information given to online
  retailers, is worth 30.49-44.62.

6
The model

• 3 types of agents a programmer, consumers, and
  firms that sell consumer products.
• The timing
• The programmer chooses shareware or adware.
• Each consumer decides whether or not to get the
  software.
• If the programmer chooses adware then firms
  choose how many ad banners to display.
• Consumers buy products and profits are realized.

7
Consumers

• Continuum of potential consumers - total mass of
  one
• Each consumer buys (at most) one software and one
  out of n consumer products (each produced by a
  different firm).
• Consumers belong to n different and equal sized
  groups
• The utility of consumers in group i is s if they
  buy product i and s-t if they buy another
  product, where s ? v-p, (p is exogenous)
• A consumer in group i learns about product i with
  prob. j with prob. 1-j, the consumer buys
  another product.
• Expected utility of a consumer who has no
  software

8
Shareware users

• Expected utility of a shareware user in group i
• Assumption 1 q ? (1-j)t
• The direct utility from software, q, is smaller
  than the expected loss from mismatch
• In equil., firms will agree to pay for ad
  banners.

9
Adware users

• The programmer can send adware users in group i
  targeted ads (impressions) about firm i.
• The prob. that a consumer in group i notices at
  least one of the ki impressions is
• The number of impressions that firm i needs to
  send to ensure prob. mi
• z? with m gt z measures of how effective ad
  banners are in attracting the attention of adware
  users.

10
Adware users

• The expected utility of adware users
• ß U0, B - the disutility from privacy
  violation
• mi(1-j)t - the expected loss from mismatch
• When z? (adware is effective), firms can get the
  same attention with fewer ads

11
Firms

• Play a role only in the adware case
• Total demand from firm i

• Nonadware users
• Info. adware users from group i
• Uninfo. users from groups other than i

12
Firms

• Firm i's objective
• p ? p - c is the per unit price-cost margin
• r is the per impression per user price charged by
  programmer

13
Programmers

• The model begins after the programmer has already
  developed the software gt the development cost is
  already sunk
• Programmer's profit from shareware is q
• Programmer's profit from adware
• In the 1st stage of the game, the programmer
  compares the profits from adware and shareware
  and decides how to distribute his software

14
Equilibrium - 3rd StageThe demand for targeted
ads

• Solving ?Pi(mi,..., mi)/?qi 0, yields for all
  i
• Since z lt 0, m(r) is a linearly decreasing demand
  function
• ? with p (products are more profitable)
• ? with z (ad banners attract more attention)
• ? with j (consumers more likely to be info.
  anyway)

15
Equilibrium - 2nd stageconsumer's demand for
adware

• A consumer will get an adware (rather than
  nothing) iff
• Only consumers with high ßs buy adware.
• Since ß U0, B, the fraction of adware users
  is

• This fraction is
• ? with the benefit from adware, qm(r)(1-j)t
• ? with the number of impressions, zln(1-m(r))

16
Equilibrium - 2nd stageconsumer's demand for
adware

• The aggregate demand of firms for ad banners is
• 3 effects of r?
• Each firm pays for fewer ad banners per adware
  user
• Fewer ads means less privacy loss so a(q,r)?
• Fewer ads means less information on products so
  a(q,r)?
• In our formulation, (i) and (ii) cancel each
  other out when (q,r) lt 1 (some consumers do not
  get software).
• r? means a(q,r)? so Q(q,r)?
• When a(q,r) 1 (all consumers get adware), only
  (i) is at work.
• r? means Q(q,r)?

17
Equilibrium 1st stageAdware or shareware?

• The programmer's profit from adware
• The programmer chooses r to maximize Oa(q,r).
• Assumption 2 B is sufficiently large
• The dist. of ß (the loss of privacy) is
  sufficiently wide.
• In equil., some consumers do not get an adware
  since their loss of privacy is too large.
• The programmer's profit from adware given r

18
Results

• Proposition 1 The solution to the programmer's
  problem
• (i) If B lt -(1-j)p/z, the programmer offers
  adware for all values of q.
• (ii) If B ? -(1-j)p/z, the programmer offers
  adware if q is small and shareware if q is large.
• Intuition
• When q is low, consumer pay little for shareware,
  but those with small ß will adopt free adware gt
  programmers make money from ad banners.
• As q?, WTP for shareware ? on the other hand,
  consumers with high ßs will never adopt adware
• Shareware may become more profitable than adware.
• Many popular software are first dist. as adware,
  but then, newer and improved versions are
  distributed as shareware (e.g., Gozilla and
  GetRight)

19
Proposition 1 shareware is provided when q is
high
20
Technological progress - z?

• Proposition 2 As z?
• The programmer adopts adware for a larger set of
  parameters and raises r.
• Both the programmer and consumers become better
  off.
• Fewer impression are sent gt less privacy loss.
• Intuition
• z? boosts the demand of firms for ad banners gt
  r?
• Adware becomes more profitable
• z? has two effects on consumers
• Holding m(r) fixed, z? means that fewer
  impressions are needed to get the same level of
  attention from users gt less privacy loss
• z? has a positive direct effect on m(r) and a
  negative indirect effect since r?. In our model,
  the two effects cancel each other out gt m(r) is
  independent of z.

21
Social efficiency

• Proposition 3 r is excessive gt Too few ad
  banners
• Intuition
• For adware users with low ß's, the benefits from
  info. exceed the loss of privacy
• For adware users with high ß's, the loss of
  privacy exceeds the benefits from info. (they buy
  the adware only because they also get a utility q
  from the software).
• As q?, there are more adware users of the second
  kind.
• But the number of these users is limited since
  when q?, the programmer switches to shareware
• the aggregate benefit of consumers from ad
  banners exceeds the associated disutility.
• The programmer fails to take into account this
  net benefit

22
Social efficiency

• Proposition 4 The programmer
• underprovides adware if rzln(1-m(r)) ? ½
• (the programmer's optimal profit when the adware
  market is covered is small likely if z is
  small),
• overprovides adware if rzln(1-m(r)) ? ½
• (the programmer's profit at the socially
  efficient price when the adware market is covered
  is large likely if z is large).

23
Bans on adware

• U.S. legislators consider regulating or even
  banning any software that monitors usage of the
  internet and transmits info. back from a location
  (Spy Block Act S.2145 and Safeguard Against
  Privacy Invasions Act, H.R. 2929).
• The legislation is intended to combat spyware and
  malware but may also effectively make it
  impossible to distribute legitimate adware on
  line.
• Proposition 5 A ban on adware hurts both
  consumers and the programmer
• Intuition
• The programmer is worse off by revealed
  preferences
• In the shareware case, p q, so consumers get no
  surplus. In the adware case, consumers with low
  ß's get a positive surplus from the adware while
  those with high ß's do not adopt adware.

24
Competition in the software market

• 2 programmers develop software of equal quality.
• To avoid Bertrand competition, one programmer
  will distribute his software as shareware and the
  other will distribute it as adware.
• We study the following game
• The programmers simultaneously choose ps and r
• Firms decide how many ad banners to pay for.
• Each consumer chooses shareware or adware.
• Consumers buy products and profits are realized.

25
Competition in the software market

• Assumption 3 B is sufficiently large (otherwise
  all buyers get an adware even if the shareware is
  given for free)
• Lemma 1 There exists a unique Nash equilibrium,
  (p,r).
• Proposition 6 As z?, p? and r?. Moreover the
  shareware programmer benefits, the adware
  programmer becomes worse-off and consumers
  benefit.
• Proposition 7 A ban on adware hurts both
  consumers and the programmers.

26
Conclusions

• We modeled the decision of programmers between
  adware and shareware
• This choice is affected in our model by
• The quality of the software favors shareware
• The quality of ad banners technology favors
  adware
• From welfare standpoint, our model suggests that
• There are two few ad banners in equilibrium
• There is too much or too little adware relative
  to shareware
• Bans on adware may be a bad idea