![](bg6.png)
ship (71%), followed by singles (24%) and divorced
or separated individuals (5%).
3.2 Measures
The online survey consisted of three main parts fol-
lowing the introduction explaining the purpose of the
data collection:
1. Basic demographic information (age, sex, level of
education, nationality).
2. Behavioral responses for deriving individuals’ de-
lay discounting rates.
3. Publicly observable features linked to the individ-
uals.
3.2.1 Delay Discounting - MCQ-21
The validated 21-item Monetary Choice Question-
naire (MCQ-21) instrument was used for collecting
responses from participants to compute each individ-
ual’s overall discounting factor k. The MCQ-21 is a
self-reporting questionnaire comprising of a set of 21
questions requiring participants to make a choice be-
tween a smaller, immediate reward (SIR) or a larger,
delayed reward (LDR) with monetary values (Kaplan,
2016). The original instructions for the question-
naire: “For each of the next 21 choices, please in-
dicate which reward you would prefer: the smaller
reward tonight, or the larger reward in the specified
number of days. Although you will not actually re-
ceive any of the money, pretend that you will actually
be receiving the amount that you indicate. Therefore,
please answer each question honestly and as if you
will actually receive the amount chosen either tonight
or after a specified number of days. To indicate your
choice, please clearly circle the amount and time as
shown in following example: 0. Would you prefer
$100 tonight, or $100 in 45 days?” (Kaplan, 2016)
were modified so they suit better for the online survey
format. For each question two radio buttons were pro-
vided to make the choice task clear: e.g. $30 tonight
or $85 in 14 days.
Discounting metrics were computed for each re-
spondent using the Excel-based automated scoring
tool, which facilitates the complex computations to
derive the discounting factor k from MCQ-21 (Ka-
plan et al., 2016). The tool reports summary statistics
for the whole sample, checks consistency and outputs
several discounting metrics on the individual level:
overall k, small k, medium k, large k, geomean k (tak-
ing the geometric mean of the small, medium, and
large k values), as well as the log and ln for each of
the k scores. The following analyses use the "overall k
factor" measuring the daily rate at which rewards lose
their value. Rearranging the equation of the hyper-
bolic function gives the formula for the discounting
factor k:
k =
A
V
− 1
D
where V is the smaller, immediate reward; A is the
larger, delayed reward; and D is the delay associated
with A. For a more detailed explanation on deriving
the overall k factor see: (Kaplan et al., 2016).
3.2.2 Publicly Observable Attributes
This section of the questionnaire aimed at collecting
information linked to respondents, which can be eas-
ily observed in most public settings (e.g. work) with-
out direct interaction with the stakeholder. Two cat-
egories of data can be distinguished: ownership of
items and habits. Ownership questions focused on
the presence of attributes, while questions related to
habits were concerned with the frequency of various
actions.
A single choice response format was used to as-
sess the presence of the attributes, and for certain
attributes, additional questions were included to ob-
tain a more detailed description. Question cate-
gories were as follows: real estate (number, loca-
tion, size), car (number, brand, model, type, color, en-
ergy source, unique license plate), motorcycle (num-
ber, brand, type), bicycle (brand, type), boat (brand,
type), phone (brand, model, color, cover, cover color),
laptop (brand, OS, size, camera cover, decoration),
tablet (brand, size), watch (type, brand), headphones
(brand), sunglasses (brand), backpack (brand), brief-
case (brand), jewellery (type, material), wallet (ma-
terial), sport equipment (17 items), pets (7 species
+ other), style description (15 categories), cosmetic
surgery, hair dye, hair length, facial hair, dietary
lifestyle (7 categories), tattoo (general categories,
place of tattoo), social media (existing accounts), pre-
ferred browser, preferred search engine.
Questions related to habits asked the frequency of
various activities on a 9-point response format where
each point had a textual label ranging from 0 - never in
the last 12 months to 8 - every day or nearly every day.
Questions assessed the frequency of: wearing certain
clothes (23 items), doing various sports (17 sports),
listening to music (14 genres), consuming drinks (11
drink types), consuming other products (6 items), en-
gaging in various other activities (26 activities).
4 RESULTS
The final dataset contained valid responses from a
total 331 subjects. The key dependent variable for
Inferring Delay Discounting Factors from Public Observables: Applications in Risk Analysis and the Design of Adaptive Incentives
75