Clarifying
Your Values and Measuring Tradeoffs:
Multi-Attribute
Value Analysis
In multiple-issue negotiations, one faces difficult
tradeoffs between issues that are qualitatively different, but each highly
valued. One must decide how much of one
issue to give up in order to obtain a gain in another issue. This exercise is designed to lead you
through some steps of qualitative and quantitative analysis that will prepare
you to make these decisions.
The field of decision analysis has developed a method for
sorting out one's preferences in situations involving multiple issues. Most negotiators find that use of this
method during preparation for a negotiation yields a substantial payoff. It increases one's confidence in tradeoff
decisions, increases the clarity with which one can communicate, and decreases
the chances of making a decision one will later regret.
Because money is quantifiable and familiar, it would be
convenient to try to convert all negotiation outcomes to dollars. However, many things of value to you are
difficult to translate into money terms.
In a job decision, can you precisely value health benefits, vacation
days, office space, computer support, job prestige, company prestige, and
industry prestige? Moreover, although
we are more comfortable thinking in terms of dollars, what we are maximizing in
life is some underlying utility function, for which money is just one way to
meet the needs and goals that give us utility.
Finally, there is often not a linear relationship between money and
utility-- money typically has diminishing marginal value. Thus, the ability to quantify underlying
values is a useful general tool for decision making and negotiation that can
supplement a purely financial analysis.
Step 1: Identify the relevant issues and range of
possible outcomes
The issues in a negotiation are not always obvious and
need to be identified by considering how the upcoming negotiation could
contribute to meeting one's interests.
List the issues that will arise in your negotiation. For each issue, list the possible outcomes
(e.g., for salary, it would be specific dollar amounts; for location, it would
be specific cities). You should list
the outcome levels that could be part of a possible settlement; that is, not a
level which is so good as to be implausible or a level so bad as to be, by
itself, a deal breaker. Carefully
identifying the range of outcome levels on an issue makes it possible to
precisely weight issues.
Example
Cathy expects to receive job offers
from three companies. After careful
reflection, Cathy decides that her main interests in the next few years will be
to live in a city with mild winters and many outdoor activities, a comfortable
salary, and large blocks of vacation time to take extended hiking trips.
All the companies pursuing Cathy
have offices in Boston, Detroit, San Francisco, and Seattle. The salary in her field ranges from $50,000
to $70,000. Three weeks vacation is typical,
but it can range from 2 weeks to 6 weeks.
Step 2: Identify the best and worst outcomes within
each issue.
Within each issue, decide which outcome is the best and
which is the worst.
Example
Cathy decides that the worst
location is Detroit and the best is San Francisco. Naturally, the worst salary is $50,000 and the best is
$70,000. The worst vacation time is 2
weeks and the best is 6 weeks.
Step 3: Assign value levels within each issue.
Assign the best outcome for a given issue the value of
100 and the worst outcome a value of 0.
Note that 0 does not mean that an outcome has no value! Rather, it means that within the range of
outcomes you have identified, it has the lowest value. Similarly, 100 does not imply perfection; it
only implies that it is the best outcome in the range identified. (Technical aside: Using 0 and 100 improves consistency when weights are assigned in
Steps 4 and 5.) Then for the other
outcomes on that issue, assign a number between 0 and 100 that reflects their
value relative to the best and worst outcomes.
Do this for all of the issues.
Example
Cathy’s most preferred city, San
Francisco, is assigned 100, and her least preferred city, Detroit, receives
0. Her assessment of outdoor
opportunities and mild winters leads her to assign Seattle a score of 80 and
New York a score of 40. She analyzes
the other issues in a similar manner.
Location Score Salary Score Vacation Score
Detroit 0 70 100 2 0
Boston 40 65 90 3 20
San Francisco 100 60 70 4 40
Seattle 80 55 40 5 70
50 0 6 100
Keep in mind, once again, that 0 does not mean that an
issue has no value to Cathy. Cathy
simply would enjoy Detroit, $50,000, and 2 weeks vacation the least.
Step 4: Assign preliminary issue weights.
Construct a package consisting of the worst outcomes on
all of your issues. For Cathy, this
would be Detroit, $50K, and 2 weeks vacation (0, 0, 0). For each issue, think about how much
enjoyment that moving from the worst outcome to the best outcome would provide.
Then look at all of the issues and decide for which one this change would have
the largest impact on your overall satisfaction. This is your top-ranked issue.
Then look for the issue on which moving from worst to best would have
the next greatest impact. This is your
second-ranked issue. Continue in this
way with all of your issues.
Now we need to assign them a weight. Assume that Cathy has ranked the issues in
the order (1) location, (2) salary, and (3) vacation. Note that ranking location ahead of salary does not mean that
location is more important than salary to Cathy in general; it simply means
that it is more important for the range
of outcomes she is considering.
There are several techniques for assigning weights. A relatively straightforward method is to
tassign your top-ranked issue the score of 1.00. Then evaluate how moving from worst to best on the second ranked
issue compares to the same movement on the first ranked issue. If it is nearly as desirable, give it a high
score (e.g., .80, or "80% as important"); if it is only about half as
desirable, give it a middle score (e.g., .50, or "50% as important"). Do this for all of the issues.
Example
Changing from the least preferred
location (Detroit) to the most preferred (San Francisco) gives Cathy the most
enjoyment. Location, therefore, is her
most important issue, and she gives it a preliminary weight of 1.00. The next most important change is from a
salary of 50,000 to a salary of 70,000, but it’s not nearly as important as
location, so she gives it a weight of .60.
Moving from 2 weeks to 6 weeks vacation is nearly as important as
salary, so she gives it a .50.
Note: An alternative method for assigning issue weights is to work from
specific tradeoffs. For example, we can
assign a weight to Cathy's second-ranked issue (W2), salary, by
comparing a job that has the best salary and the worst location (100W2
+ 0W1) to a job that has the
worst salary and a move to a better location (0W2 + xW1). How much improvement in location would be
needed to make these equivalent in value?
Would a move to Boston be enough?
If not, we know that x > 40.
Would a move to Seattle be enough?
If so, we know that x < 80.
Let's say that x does lie between 80 and 40, and Cathy splits the
difference in value, setting x = 60.
Once we know the value required on the location issue that makes Cathy
indifferent (60), and given that the weight of the location issue has been set
to 1 (W1 = 1.00), it follows that W2
= .60.
Step 5: Normalize the issue weights.
Sum the points you've assigned to all of the issues (in
the example, 1.00 + .60 + .50 = 2.10).
Divide the points you've assigned to each issue by this total (e.g.,
1.00/2.10, .60/2.10, and so on). Now
you have converted all of the points to a standard scale that sums to 1.00, and
that reflects the issues’ relative weight (e.g., Cathy’s most important issue
has a weight of .48, the next most important issue has a weight of .29, and so
on).
Step 6: Multiply the outcome values and issue
weights.
Finally, take the product of the normalized weight for
each issue (Step 5) and the value you assigned to each outcome on that issue
(Step 3). This gives you a standardized
value score for each outcome. Now all
possible combinations of outcomes can be compared to each other, and to your
ideal and worst set of outcomes (which have scores of 100 and 0, respectively).
The completed example is on the next page. As you look at it, imagine Cathy's offer
from Comapny A is (Boston, $60,000, 4 weeks vacation) and from Company B is
(Seattle, $50,000, 3 weeks vacation).
Which does she prefer? What
should she negotiate for?
Beyond Step 6:
Before relying on the MAV model you have constructed, you
should test the model in two ways. First,
look at the tradeoffs implied by the model.
For example, Cathy should ask whether an increase in salary from $55,000
to $65,000 (+14 points) would truly offset a change in assignment from San
Francisco to Seattle (-10 points).
Second, generate a number of alternative settlements and check whether
the preferences implied by the model match your direct preferences. If not, you should adjust the weights and
values until the model's preferences and your direct preferences converge.
This approach makes the assumption of additivity. In some instances, value may not be
additive. The value of an outcome on
one issue may depend on the outcome of another issue (for example, the value of
a $70,000 salary varies depending on whether you live in New York or
Boise). Usually there will be a
redefinition of issues that will allow you to use an additive system (e.g.,
instead of using salary, you can use salary adjusted for cost of living).
A final point that should be kept in mind is that the
scores you have created are designed only to prioritize and measure your own
conflicting interests. The scores
permit you to measure all outcomes on the same scale and to assess your
preferences for bundles of outcomes.
Note, however, that absolute utility scores are not meaningful to others
("San Francisco gives me 48 utiles") because the absolute scores are
the function of an arbitrarily chosen scale (100 points) that is conditioned on
your own specific set of best and worst outcomes.
Of course, constructing the MAV model on a spreadsheet is
a good idea because it allows you to answer "what if" questions when
preparing one's strategy. Also, use of
the spreadsheet during the negotiation allows for quick evaluation of
offers.
For people who like this kind of thing only: Another program that can aid this process is
one that performs a conjoint analysis which works on the logic of
"revealed preference". Those
of you have had a few marketing or statistics courses might have encountered it. With these programs, you merely have to rate
or rank a number of possible settlements, and the computer derives the weights
and values that underlie your choices. Rather
than having to work top-down to decide how much weight you place on each issue
and outcome, you can work bottom-up from your choices and let your actions tell
you what you care about!
|
Step 1 |
Step 2 |
Step 3 |
Step 4 |
Step 5 |
Step 6 |
|
Identify relevant issues
and outcomes |
Identify best and worst
outcomes for each issue |
Assign value levels for
each outcome in an issue |
Identify preliminary
weights |
Normalize the weights |
Multiply normwts by
values (st 3 x 5) |
|
Location |
|
|
1.00 |
1.00/2.10=.48 |
|
|
Boston |
|
40 |
|
|
19 |
|
Detroit |
worst |
0 |
|
|
0 |
|
SF |
best |
100 |
|
|
48 |
|
Seattle |
|
80 |
|
|
38 |
|
|
|
|
|
|
|
|
Salary |
|
|
.60 |
.60/2.10 = .29 |
|
|
$70 |
best |
100 |
|
|
29 |
|
65 |
|
90 |
|
|
26 |
|
60 |
|
70 |
|
|
20 |
|
55 |
|
40 |
|
|
12 |
|
50 |
worst |
0 |
|
|
0 |
|
|
|
|
|
|
|
|
Vacation |
|
|
.50 |
.50/2.10 = .24 |
|
|
2 wks |
worst |
0 |
|
|
0 |
|
3 wks |
|
20 |
|
|
5 |
|
4 wks |
|
40 |
|
|
10 |
|
5 wks |
|
70 |
|
|
17 |
|
6 wks |
best |
100 |
|
|
24 |
|
|
|
|
Swt = 2.10 |
Snwt = 1.00 |
|