2012 Survey
Sample Frame and Sample Design
With the advent of the
USGS
Earth Resource Observation Sci-
ence Center (
EROS
) providing free access to Landsat imagery, a
requirement was made for users to register with
EROS
in order
to download images. This provided a complete sample frame
of direct or initial users of Landsat images. We say direct or
initial users, as we are aware that there may be literally mil-
lions of subsequent users of the images produced by the ini-
tial users (e.g., users of Google
™
maps,
GPS
programs). Thus,
our estimates of the benefits of Landsat images are conserva-
tive as only the initial direct users were sampled.
The sample frame was 44,731 unique working email ad-
dresses of registered Landsat users worldwide. Because the
survey was being conducted online, we chose to do a census
of all the users rather than a sample. A variety of questions
were asked of the respondents in addition to the
WTP
ques-
tions, including questions related to the specific use and
qualitative value of satellite imagery. Additional questions
were used to elicit conditioning information with respect to
the respondents. The respondents are direct users registered
with
EROS
, and information was sought so as to be able to
describe how the value of imagery varied across categories of
these direct users.
Survey Implementation
The survey was launched in April of 2012. The implementa-
tion followed the Dillman (2007) repeat contact method to at-
tain as high a survey response rate as feasible. A total of 13,473
individuals responded to the survey, for a response rate of 30
percent. This response rate is equivalent to the average re-
sponse rates for online surveys reported in several meta-anal-
yses (for example, Lozar Manfreda, 2008; Sheehan, 2001; Shih
and Fan, 2008). This number includes both completed surveys
(
n
= 11,749) and partially completed surveys (
n
= 1,724).
A non-response survey was conducted among those who
did not respond to the initial survey to ascertain if those who
did not respond were different than those who did. This short
survey contained four questions: current or past Landsat user;
new, established, or returning user; sector (government, pri-
vate sector, academia, etc.); and citizenship. A total of 1,622
individuals responded to the non-response survey. These
results were compared to those of the respondent sample, and
it was determined that there were no statistically significant
differences in the distribution of the four variables between
respondents and non-respondents.
Willingness-to-Pay Question
The
CVM
question in the 2012 survey was very similar to that
in the 2009 survey in that it was a hypothetical but realistic
scenario and included an explicit budget restriction. At the
time the 2012 survey was administered, Landsat-5, one of the
two satellites operating at the time, had stopped acquiring
imagery. Some users had already replaced Landsat-5 imagery
with data from other sources, so they had recent experience
with paying for replacement information. The specific ques-
tion asked on the survey was:
“At the moment, current Landsat 5 imagery is not available
(expected to be available again in spring of 2012) and you
may have already obtained imagery elsewhere to replace
Landsat 5. If both Landsat 5 and 7 became permanently
inoperable before the next Landsat satellite is operational
(scheduled to launch in early 2013), you may have to
obtain imagery elsewhere again. Assume that you are re-
stricted to your current project or agency budget level and
that the money to pay this cost would have to come out
of your existing budget. If such a break in continuity did
occur and you had to pay for imagery that was equivalent
to the Landsat standard product typically available (which
assumes both Landsat 5 and 7 imagery are available),
would you pay $X for one scene covering the area equiva-
lent to a Landsat scene?”
The “$X” was randomly filled in with one of 20 different
dollar, or bid, amounts, ranging from a low of $10 to a high
of $20,000 (this range was set after the 2009 pilot survey).
Respondents were instructed to answer
Yes
or
No
.
Some assessment of the downstream value of Landsat im-
agery is implicit but not the main focus of the question. The
question asks the user whether they would pay $X to replace
the lost imagery. Thus, the respondent values the information
based on a comparison of cost or lost revenue of replacing the
imagery. For example, if a for-profit firm sells an interpretive
service of Landsat imagery for $Y and has an in-house cost of
$Z for developing that service, then presumably the business
user would be willing to pay some portion of $(Y − Z) for
alternative imagery. Likewise, if a government agency is man-
dated to provide a service, such as reporting crop forecasts,
they must obtain information to fulfill that requirement. If
they have a budget of $A for the use of satellite imagery and
replacing the information provided by the imagery would cost
$B, the government user would presumably be willing to pay
up to $B.
After the first
WTP
question was asked and answered,
a follow-up second question was asked. If the respondent
answered
Yes
to the first question, they were asked if they
would pay a second higher bid amount. If the respondent an-
swered
No
to the first question, they were asked if they would
pay a second lower bid amount. This follow-up question
provides more information about the respondent’s underlying
WTP
and enables more efficient estimation of the distribution
of
WTP
across users (Hanemann
et al
., 1991). Lastly, if the
respondent answered
No
/
No
then a final bid amount of $1
per scene was asked. An answer of
No
to this trivial amount
indicates a protest bid; that is, the respondent is not willing
to reveal their valuation of the imagery and may be offended
by being asked to value the information hypothetically. As is
customary, protest bidders are not included in the statistical
analysis (Mitchell and Carson, 1988).
Statistical Analysis
Using the results of the
CVM
question, a measure of economic
benefits received by direct users for the equivalent of a Land-
sat scene can be obtained. Responses to the first
CVM
ques-
tion provide the data necessary to estimate a single-bounded
estimate of
WTP
. The underlying distribution of
WTP
, which is
unknown, can be specified as:
WTP
i
* =
x
i
´
δ
+
ε
i
(1)
where
x
i
is a vector of independent variables, including the
dollar bid amount, that are expected to influence individual
i
’s willingness-to-pay for the imagery,
δ
are a vector of weights
associated with these attributes, and
ε
i
is a random error term.
This dollar amount varies across the sample to provide in-
sights about the distribution that representative respondents
would pay. Whether or not an individual is willing to pay
a particular bid amount is observed, so the probability that
individual
i
responds
Yes
to a specified bid amount, bid
i
, is
equal to the probability that the random willingness-to-pay
function is greater than or equal to that bid amount:
Pr(WTP
i
*
≥
bid
i
|
x
i
) = Pr(Y = 1|
x
i
) = 1 –
F
(bid
i
|
x
i
) (2)
where
F
is a cumulative distribution function (
CDF
). Likewise,
there is a similar relationship for individuals responding
No
to their specific bid amount:
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
August 2015
649
