PE&RS August 2015 - page 651

use of the imagery. Similarly, other bid amounts can be used
to determine the proportion of users that would forgo use of
the imagery if those amounts were charged. Finally, the mean
value of a scene to a group of representative users can be mul-
tiplied by the number of scenes obtained by that group over a
given time period and then aggregated across all groups.
Results
Descriptive Results
The majority of respondents were current Landsat users (84
percent; users who had used Landsat in their work in the
year prior to the survey) and the remainder were past Landsat
users (16 percent). The results reported here apply to cur-
rent Landsat users registered with USGS EROS. We accounted
for three variables within the analyses: sector, residence, and
type of user. Based on the results from the 2009 pilot survey,
we expected to find differences in WTP by sector. The majority
of users in the 2012 survey were from academic institutions
(slightly more than 50 percent), and about 15 percent of the
users were each from private business and Federal govern-
ment. About 5 percent were each from state/local government
and non-profit organizations. Additionally, we wanted to
separate US and international users to note any WTP differenc-
es between those two groups, especially since US users down-
load the majority of scenes from USGS. Slightly more than
one-quarter (27 percent) of the current Landsat users were US
residents and 73 percent were from other countries (inter-
national users). Finally, we divided users into three types:
established (those who used Landsat regularly both before
and after it became available at no cost), new (those who had
never used Landsat imagery before it became available at no
cost), and returning (those who had used Landsat in the past,
but had not used Landsat for at least a year prior to it becom-
ing available at no cost). We hypothesized that established us-
ers would report greater benefits from using Landsat imagery
than new or returning users, based on their consistent use of
the imagery over time. New users were the largest group of
users (43 percent), followed by established users (41 percent),
and returning users (16 percent). In the analysis, new and
returning users were combined due to their similarities across
a variety of variables. Based on these variables, we report re-
sults for four groups of users: (a) US established users, (b) US
new/returning users, (c) international established users, and
(d) international new/returning users. The variation in WTP by
sector is also taken into account within each of these groups.
Economic Valuation Results
Statistical Results
Table 1 presents the dichotomous choice double-bounded logit
valuation model results. There is a separate logit regression
model for each of the four groups: US established users, US
new/returning users, international established users, and in-
ternational new/returning users. Within each group, the model
incorporates a variable in the form of dummy intercept shifter
variables for the type of sector to which the user belonged:
academic, Federal government, state and local government,
private for-profit business, and non-governmental organiza-
tions. These dummy or shifter variables information results in
a separate intercept parameter for each group. Changes in the
slope coefficient by group were also tested for and found to
not be statistically significant. Within each of the four mod-
els, separate valuations are allowed for each of the user types
since the intercepts strongly differ across user types within
each group. Slope parameters differ strongly across established
versus new/returning and US versus international groups.
As can be seen in Table 1, the sign coefficient for the dollar
bid amount the user was asked to pay is negative and sig-
nificant for each of the user groups. This indicates internal
validity of results, i.e., the higher the dollar amount (bid) a
user is asked to pay, the lower the probability they will pay
this amount to download a scene. This intuitive pattern ex-
hibits economic behavior consistent with economists’ law of
T able 1. 2012 Landsat User Survey Results from Double-Bounded Dichotomous Choice Models 1
U.S. Users
Established Users New/Returning Users
Estimate Std Err Estimate Std Err
Intercept 4.3041*** 0.2630 3.4689*** 0.2066
$ Bid Slope -0.8788*** 0.0411 -0.9071*** 0.0374
Federal Gov Shifter 0.6779*** 0.1871 0.4685** 0.2100
State & Local Gov Shifter 0.4517 0.3084 0.2790 0.2755
Private Business Shifter 0.3868** 0.1964 0.4014** 0.1876
Non Gov Organizations
Shifter 0.9066** 0.3771 0.3738 0.3244
% Correct Predictions 76.5% 85.8%
International Users
Established Users New/Returning Users
Estimate Std Err Estimate Std Err
Intercept 4.1325*** 0.1323 3.1537*** 0.1165
$ Bid Slope -0.8475*** 0.0216 -0.8030*** 0.0207
Federal Gov Shifter 0.4563*** 0.1433 0.4694*** 0.1574
State & Local Gov Shifter 0.6218*** 0.1634 0.3710** 0.1770
Private Business Shifter 0.7042*** 0.1134 0.6317*** 0.1228
Non Gov Organizations
Shifter 0.1196 1.1626 0.2144 0.1805
% Correct Predictions 74.4% 81.5%
1Academic sector is implicit within the intercept and all sectors shifter coefficients are measured relative to academics (i.e., positive signs on
the shifter variables indicate a higher value than academics).
*, ** and *** indicate statistically significant at the 10 percent, 5 percent and 1 percent level.
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING August 2015 651