Escolar Documentos
Profissional Documentos
Cultura Documentos
Step 3 – Orient the respondent. Introduce the study by an orientation page (see Fi
gure 2). The orientation does not tell the respondent more than is absolutely ne
cessary.
Figure 2 – Example of the orientation page. This particular study dealt with terro
rism.
Step 4 – Run the actual evaluation of test concepts. Present each respondent with
different combinations of elements, according to a ‘main effects’ experimental desig
n. With four silos, some concepts comprise one element from each silo, whereas o
ther concepts comprise one element from three of the four silos. Figure 3 shows
an example of a test concept. Again, keep in mind that each respondent evaluates
different sets of concepts. It is virtually impossible for a respondent to keep
track of his answers or trace those answers back to the specific element. As th
e interview progresses, the respondent is forced to answer questions at a rapid
pace and, frequently, does not stop to give rational answers. It is important to
make sure that the question is ‘meaningful’, intuitively easy to answer for the top
ic, and anchored at both ends of the scale, in order to reduce confusion and amb
iguity.
Figure 3 – Example of a test concept
Step 5 – Create a model for each respondent. The independent variables in the mode
l are the 36 elements, which are present or absent in the concept. These are so-
called ‘dummy variables’. The dependent variable is either the rating itself or, mor
e frequently, a binary conversation of the variable. For this database, the rati
ngs 1-6 (easy to deal with the issue) were converted to 0, whereas the ratings 7
-9 (hard to deal with the issue) were converted to 100. For each respondent rela
te the presence/absence of the 36 concept elements to the ratings, using the met
hod of ordinary least squares. Each dependent variable generates its own equatio
n, of the form:
Rating = k0 + k1(Element 1) +
k2 (Element 2) .. k36(Element36)
Step 6 – Segment the respondents by the pattern of their 36 ‘utility’ or impact values
. The segmentation used so-called k-means clustering (Systat, 1997). We identifi
ed two meaningful clusters or segments, although there might possibly be more.
Step 7 – Look at the additive constant across the 15 studies in the database. The
additive constant tells us the basic level of the respondents’ inability to deal w
ith the situation. Respondents do not estimate this basic level of inability. Ra
ther, it is deduced from the pattern of responses. Regression analysis partials
out the contributions of the individual elements to a person’s rated inability to
deal with the situation. Once we partial out those contributions, what is left i
s the respondents’ basic inability. Respondents cannot consciously generate a pre
determined level of base anxiety from the pattern of responses. In Figure 4, w
e see that personal situations such as income are far more likely to generate a
basic feeling of ‘I cannot deal with this’ compared to more general problems such as
terrorism. Indeed, terrorism is on par with phobias in terms of the respondents’
basic inability to deal, even though it would be counter-intuitive if this conc
ept was asked in a regular survey!
Figure 4 – Ability to cope with a situation. The bars represent the additive const
ant for the total panel. The additive constant shows the basic level of the resp
ondents’ inability to cope with a situation, in the absence of any messaging.
Step 8 – Look at the ‘knowns’ to establish consistency and validity. One way to estab
lish that the survey is valid is to look at the relationship between the indepen
dent variables and dependent variables. For example, we know that a yellow alert
signifies a less serious threat than an orange alert, which, in turn, signifie
s a less serious threat than the red alert. When these alerts are embedded in a
set of concepts with everything else in the concept also varying, do we see thi
s order of red>orange>yellow in terms of being unable to deal with the issue? T
he answer is yes, although the effect is marginal. Looking at Table 2 (below), w
e see the impact or utility value. A negative number means relief (pushes away f
rom; can’t deal with the situation), whereas a positive number means increasing a
nxiety. The order of colors is exactly what we expect. Yellow is less anxiety-p
rovoking than orange, which is less anxiety-provoking than red. What is just as
important is that the effects are minimal. The dynamics for war are slightly dif
ferent from those of terrorism. A red alert during war is almost irrelevant, whe
reas a yellow alert is a cause of a bit of relief. Both are anxiety-provoking fo
r terrorism, as they should be.
Table 2 – How different alert colors drive ability to deal with the situation (neg
ative numbers) or drive inability to deal with the situation (positive numbers)
War Terrorism
B7 During a Yellow alert… -2 2
B8 During an Orange alert… -1 2
B9 During a Red alert… 1 4
Step 9 – Look at what is most anxiety-provoking versus most anxiety-reducing. Most
of the elements had little effect. However, some interesting repeating patterns
and surprises emerge. Direct personal threats are clearly very anxiety-provokin
g and, more so, for terrorism than for war. But, surprisingly, so is the involve
ment of the United Nations. This is an exceptionally anxiety-provoking statement
, since it is the opposite of the original concept of what the UN was originally
created to do.
In the context of terrorism, a computer virus elicits a marginal response, provi
ding a bit of comic relief, except, perhaps, to a person whose computer is infec
ted with the virus. God, friends, family, and information provided by the media
are all anxiety-reducers. It appears that the basic ability of respondents to de
aling with the concepts of terrorism and war are equal. After that, however, spe
cifics for terrorism generate greater anxiety than those same specifics for war.
(See Table 3).
Table 3: The most impactful elements driving the ability to deal with a situatio
n (negative numbers) or the inability to deal with a situation (positive numbers
).
War War Terrorism
Base Size 124 Base Size 121
Constant 20 Constant 19
A9 A dirty nuclear bomb set off… 16 A3 A bomb under you
r car… 21
A6 Bombs blowing up in the middle of a city… 9 A9
A dirty nuclear bomb set off … 20
A8 A deadly disease like smallpox or anthrax let loose... 9
D2 You believe that international cooperation in the United Nations
will keep you safe 20
D3 You believe the United Nations Forces will keep you safe 8
D3 You think United Nations Forces will keep you safe 19
D2 You believe international cooperation in the United Nations will keep yo
u safe 8 A4 Bombs blowing up in the middle of a buil
ding… 15
A7 A deadly disease like smallpox or anthra
x let loose... 14
D5 You believe that the Center for Disease
Control will keep you safe 12
A6 Contamination of the food supply…
10
D4 You believe that Homeland Defense will k
eep you safe 10
C8 Family and Friends play a big role in your life… -3
C6 You trust that God will keep you safe -3
D9 You need to contact your friends and family to make sure they are OK…
-4 C5 It’s important for the Media will keep you informed
-5
A4 A build up to war that ends up a false alarm… -6 B8
You need to contact your friends and family to make sure they are OK… -6
D1 You trust that God will keep you safe -6
D8 It’s important for the Media to keep you informed -7
10. Segment the respondents by the pattern of their utilities. Segmentation prov
ides a way of delving into mind-sets. Public opinion and market research studies
often segment responses on the basis of patterns of reactions to questions that
ask respondents to ‘define themselves’. Our goal, in this data set, is to generate
segmentation based upon patterns of reactions to the concept elements. Since the
respondents do not tell the interviewer what is important, the importance or dr
iving power of the individual elements is obtained from the pattern of responses
.
We see these patterns when we divide the respondents by the patterns of their ut
ilities;
War (Table 4) – Two segments emerge. The bigger group is the ‘Frightened Ordinary Ci
tizen’, the smaller group is the ‘Local Patriot. The local patriot is less anxiety p
rone (constant = 15) than the frightened ordinary citizen (constant = 23). The
Frightened Ordinary Citizen is afraid of injuries, and is comforted by religion.
The Local Patriot is afraid of involvement of international bodies or even the
government in general, and is comforted by seeing friends become comrades in arm
s.
Table 4: War (total and segments). The table shows the most impactful elements d
riving the ability to deal with a situation (large negative numbers) and the mos
t impactful elements driving the inability to deal with a situation (large posit
ive numbers).
War
Tot FOC LP
Base Size 124 77 47
Constant 20 23 15
‘Frightened Ordinary Citizen’
A dirty nuclear bomb set off… 16 23 5
A deadly disease like smallpox or anthrax let loose... 9 16
-3
Bombs blowing up in the middle of a city… 9 14 0
Friendly Fire killing troops... 5 10 -2
Seeing my friends or family getting called up to go fight… 3 8
-5
A build up to war that ends up a false alarm… -6 -9 -2
Its important for the Media to keep you informed -7 -9 -2
You trust that God will keep you safe -6 -14 7
‘Local Patriot ’
You believe international cooperation in the United Nations will keep you safe
8 -4 28
You believe the United Nations Forces will keep you safe 8 -3
27
You believe the National Reserves will keep you safe 2 -6 14
You trust your Local police will keep you safe 1 -6 13
You trust your Local hospital will keep you safe 3 -3 12
You believe the Military will keep you safe 1 -5 10
During a Red alert… 1 3 -3
A deadly disease like smallpox or anthrax let loose... 9 16
-3
In a non populated area… -3 -3 -3
An area crowded with soldiers… 0 2 -3
Seeing my friends or family getting called up to go fight… 3 8
-5
Terrorism (Table 5) – The same two segments emerge, based upon the pattern of util
ity values. However, the dynamics differ. The Frightened Ordinary Citizens are t
he slightly smaller group (N=57 versus N=67 for the Local Patriot). Furthermore,
the Frightened Ordinary Citizen is less immediately plagued by general anxiety
(additive constant = 9) than is the Local Patriot (additive constant = 27).
However, the dynamics of the elements are similar to what we saw for war, only m
ore pronounced. Both groups are much more anxiety-prone to what bothers them, w
ith impact values of 30 or more for a number of situations. Very little action w
ill reduce the anxiety of the Frightened Ordinary Citizen. The computer virus ag
ain is perhaps a bit of comic relief, when juxtaposed against elements that deal
with dirty nuclear bombs, or bombs under one’s car. In contrast, the Local Patri
ot is anxious about the involvement of the authorities, no matter who they are,
and finds anxiety relief with a personal, and the knowledge that friends and fam
ily are safe.
Table 5: Terrorism (total and segments). The table shows the most impactful elem
ents driving the ability to deal with a situation (large negative numbers) and t
he most impactful elements driving the inability to deal with a situation (large
positive numbers).
Terrorism
Tot FOC LP
Base Size 121 54 67
Constant 19 9 27
‘Frightened Ordinary Citizen’
A dirty nuclear bomb set off … 20 39 5
A bomb under your car… 21 38 8
Bombs blowing up in the middle of a building… 15 31 2
A deadly disease like smallpox or anthrax let loose... 14 28
3
Contamination of the food supply… 10 19 3
An area crowded with children… 6 11 2
In a heavily populated area… 3 10 -3
Fire raging through a building… 4 9 0
A Computer virus let loose that impacts your everyday businesses… -3
-4 -2
‘Local Patriot ’
You think United Nations Forces will keep you safe 19 0 34
You believe that international cooperation in the United Nations will keep you s
afe 20 3 33
You believe that Homeland Defense will keep you safe 10 -3 20
You believe that the Center for Disease Control will keep you safe 12
3 18
You think that your Local police will keep you safe 7 -1 14
You think that your Local hospital will keep you safe 6 -1 12
A bomb under your car… 21 38 8
An area filled with tourists… 1 8 -4
At a turning point in your life... -2 1 -4
In a non-populated area… -1 3 -4
The media talking about potential terrorism acts… -2 2 -5
Family and Friends play a big role in your life… -2 1 -5
You d drive any distance to get away from it… -2 5 -7
Its important for the Media will keep you informed -5 -2 -7
You trust that God will keep you safe -3 5 -10
You need to contact your friends and family to make sure they are OK… -6
-1 -11
Some observations on world-view, method, analytic strategies
Observations on surveys versus experimental design
It is worth pointing out that this study was an experiment to identify h
ow respondents integrate information about different aspects of an anxiety-prov
oking situation, weight the impact of the information, and then come up with a s
ingle answer as dictated by the research instructions. It is also important to k
eep in mind that the approach owes a lot to experimental psychology, with its fo
cus on systematically varying the situation (test concepts), measuring reactions
(ratings), and then searching for a pattern of responses. The critical differen
ce between this type of work and conventional surveys is that surveys are not e
xperiments, but just means by which to elicit the attitudes of the respondents.
Here, the focus is on actual behavior of respondents when they are confronted wi
th a collection of survey issues.
Benefits of experimental design
The world of experimental science always seems to fight with the world o
f observational science. Both disciplines have their place in the pantheon of th
e sciences that make up our world. However, they are different. Conventional sur
vey-taking usually works with the goal of measuring what is, and only occasional
ly works as a measure before or after an experimental intervention. In a sense,
the researcher contributes a great deal in survey research. It is the astute res
earcher who can draw out insights from a set of data, who sees patterns and rela
tions that were not obvious. The greater the number of years that the observatio
nal researcher works, the more likely that the researcher will add greater value
to the research process, much like the experienced clinician adds value.
Perhaps the same can be said for the experimenter, but we are not really
sure.
The experimenter is skilled in setting up experiments. In some rare cases, the e
xperimental design might require a talented experimenter. Yet, for the most part
, most really do not. Experimental science is open, public and, for the most pa
rt, available to all researchers. Perhaps the key benefit for the experimental r
esearcher is that the discipline of experimental design forces the researcher to
create the conditions to discover something. That is far more possible than wha
t confronts the observation researcher. That researcher must wait until somethin
g ‘pops out’, perhaps totally unexpectedly, and then be sufficiently aware of the im
portance of the phenomenon to jump on it, and present it in all its facets.
Acknowledgment
Dr. Howard R. Moskowitz wishes to thank his editorial assistant, Linda Lieberman
, for reviewing and preparing this paper and generally keeping him on production
schedule.
References
Beckley, Jacqueline, and Howard R. Moskowitz. Databasing the Consumer Mind: the
Crave It!, Drink It!, Buy It!, Protect It!, and the Healthy You! Databases. Annu
al Meeting, June 2002, Institute of Food Technologists. Anaheim, CA: Institute o
f Food Technologists, 2002.
Box, G.E.P.., J. Hunter, and S. Hunter. Statistics for Experimenters. New York:
John Wiley, 1978.
Luce, R. D., and J. W. Tukey. "Conjoint Analysis: a New Form of Fundamental Meas
urements." Journal of Mathematical Psychology 1 (1964): 1-36.
Moskowitz, Howard R., Alex Gofman, Rachel Katz, Barbara Itty, Madhu Manchaiah,
and Zhenyu Ma. "Rapid, Inexpensive, Actionable Concept Generation & Optimization
-the Use and Promise of Self-Authoring Conjoint Analysis for the Foodservice Ind
ustry." Foodservice Technology. 2001. 149-168.
Moskowitz, Howard R., and Alex Gofman. United States. Patent Office. System and
Method for Performing Conjoint Analysis. US Patent Pending, US2005/0177398 A1,
August 2005.
Moskowitz, Howard R., Sabastiano Poretta, and Matthias Silcher. Concept Researc
h in Food Product Design and Development. Ames, IA: Blackwell Professional, 2005
.
Stevens, S. S. Psychophysics: an Introduction to Its Perceptual, Neural and Soci
al Concepts. New York, NY: John Wiley, 1975.
Systat. Systat, the System for Statistics. Version 11 ed. Evanston, ILL: Systat,
2007.
Wittink, D. R., M. Vriens, and W. Burhenne. "Commercial Use of Conjoint Analysi
s in Europe: Results and Critical Reflections." International Journal of Researc
h in Marketing 11.S (1994): 41-52.