Consumer Usage of Broadband Internet Services - Adoption, Usage, and Global Impact of Broadband Technologies - page 213

11 The multinomial logit requires the distribu-
tion of the random error terms to be inde-
pendent and identical over the alternatives.
This can produce biased estimates if the
cross-elasticities between all pairs of alter-
natives are identical. The major weakness
of the multinomial logit is that the choice
between alternatives (i.e., DSL versus cable)
depends solely on the characteristics of those
alternatives being compared, excluding the
characteristics of any other alternatives pos-
sible.
12 It is unclear from the data why the pres-
ence of children and larger household size
decreases the probability of having Internet
access given the common understanding that
Internet access is useful for education and
entertainment. We note that while “study
search”, which would appear to be an educa-
tional use, is a prominent use of the Internet in
Portugal, general information searching and
obtaining news are comparable and might
not be related to children or household size.
This issue would appear to require further
study to find an answer in which we could
place more confidence.

13 The actual coefficients are -0.7594721*dc
_3+0.0799416*dc_3 2 .
14 The minimum is 0.7594721/
(2*0.0799416)=4.75 household members.
15 IDC, as cited by Standard and Poor's (2009).
16 Result is not shown in the figure.
17 Results are provided in the Appendix, sec-
tion 1b. In the 2008 ICSCE survey we do
not have a question asking hours of use; Q92
asks “How often do you usually use Internet
in your home?” with possible responses of:
several times a day, once a day, two to three
times a week, and less than once a week.
The 2008 estimation is therefore an ordered
logit instead of a negative binomial model
as used for the 2006 analysis.
18 Respondents are asked what type of home
Internet access they use, and those respond-
ing “nomadic” may therefore interpret this
question as asking what type of Internet
access they use either most frequently for
personal reasons, or access other than at and
for work.
19 For more on these and other examples of
societal impacts, see Jamison (2007).