Measures
The purpose of our empirical investigation is to determine
the extent to which UE affects sales in the
presence of buyer UA and controls for household income.
The dependent variable and proxy for UT in
this investigation is an invitee's dollar expenditure at
a home party. The amount an invitee spends results
from a decision to use a portion of her income to acquire
goods {UA) and to foster good will {UE). We assume
therefore that invitees who are not attracted by
the UA and UE they can obtain from a home party are
less likely to attend a party and purchase. Since virtually
all invitees who attend parties ultimately purchase,
invitee expenditures capture both attendance
and purchase behavior. Measures for the independent
and covariate variables are discussed in the following
sections.
Measures OJUE. Like acquisition utility, exchange
utility cannot be measured directly, but its presence
varies with (1) the strength ofthe social ties that bind
invitees with hostesses and demonstrators and (2) the
extent an invitee is previously obligated to her hostess.
These indicators capture two aspects ofthe inherent
value of a social tie to its members (i.e., its UE).
The first dimension, the enduring strength of a tie,
determines the social capital available. The greater
the social capital, the greater the returns on investment
and the greater the UE a buyer can obtain by
making an incremental investment in a social tie.
The second dimension of UE involves the degree
one partner is currently obligated to the other. In the
short term, members of dyads alternate between the
status of debtor and creditor as a result ofthe give and
take of social exchange. Norms of reciprocity motivate
the partner who is the net debtor to reciprocate
outstanding favors before requesting additional favors.
We thus expect the UE in a social tie to fluctuate
according to the level of outstanding obligation. In
other words, when one partner is obligated to another,
she is more likely to purchase and to spend
more on the purchase. Measures for these two UE constructs
are described below.
TieStrength. Granovetter (1973, p. 1361) defined
tie strength as "a (probably linear) combination ofthe
amount of time, the emotional intensity, the intimacy
(mutual confiding) and the reciprocal services
which characterize the tie." Marsden and Campbell
(1984) identify two types of tie strength measures: indicators
and predictors. Indicators are components of
pany. Further, there were no significant differences between the sociometric
characteristics of respondents and nonrespondents. Neither
were there significant differences in the demographic and
consumption characteristics of women who returned questionnaires
in response to our first request and those who responded only
after repeated requests.
tie Strength such as closeness or supportiveness. Predictors
are aspects of relationships that are related to
tie strength without being actual components of it,
such as the role content of the relationship or the
number of common, overlapping organizational
memberships. Their analysis shows that indicators
are more reliable than predictors.
Data were gathered for four indicators of tie
strength: closeness, intimacy, support, and association.
Respondents, those members of relational dyads
who provided data on social ties, were asked to describe
their relations to a set of "target" individuals
named on the stimulus list. A target individual was
the remaining member of a dyad.
Marsden and Campbell's preferred indicator,
closeness, was measured directly by asking each respondent
to rate her relationship to a target individual
on a 10-point scale (where 10 = "extraordinarily
close" and 1 = "not close at all"). The remaining indicators
were based on respondents' reported likelihood
of engaging in various activities with a respondent.
A six-point scale (where 5 = "very likely," 4
= "likely," 3 = "neither likely nor unlikely," 2
= "unlikely," 1 = "very unlikely," and 0 = "I don't
know this person") provided a measure of this likelihood.
Intimacy was measured by a respondent's reported
likelihood of sharing a personal confidence with a target
individual. Support was measured by a respondent's
reported likelihood of extending everyday (as
opposed to emergency) assistance to a target individual
(Wellman 1979; Wellman etal. 1987). Finally, association
was measured by a respondent's professed
likelihood of spending a free afternoon with a target
individual. The four indicators were then summed to
create a 25-point, composite tie strength measure.
Composite measures were created for both inviteehostess
(IH) and invitee-demonstrator (ID) ties.
Reliability estimates were calculated to evaluate
the internal consistency and the longitudinal stability
of the tie strength composite. Cronbach's Alpha for
the tie strength composite was .93 for both the preand
post-party samples. The test-retest correlation
coefficient was .91 after a three-week interval between
pre- and post-party observations. Finally, there was
a .65 correlation between respondent and target tie
strength estimates, indicating a high degree of consensus
within a dyad regarding the status of a tie.^
Invitee Obligation. Data used to measure the degree
an invitee was obligated to her hostess prior to
the party were collected from invitee responses to two
questions. The first question asked invitees to indicate
whether the hostess owed the invitee a favor or
vice versa. Invitees completed a five-point scale that
'The tie strength components were also factor analyzed. They all
loaded evenly and heavily on the first factor.
EMBEDDED MARKETS
described the extent of general obligation present in
the tie. The scale values were: -2 = "she definitely
owes me," -I = "she probably owes me," 0 = "it's
about equal," +1 = "I probably owe her," +2 = "I
definitely owe her." "I don't know this person" was
coded 0 to indicate no obligation present. Positive
numbers on this scale indicate a perceived social debt
by the invitee; negative ones indicate creditor status.
The second question asked invitees to describe
market-specific obligations incurred within the home
party market. Invitees were asked to describe (1) the
number of parties an invitee had hosted that were attended
by the present hostess of the sampled party
and (2) the number of parties the invitee had attended
in the past that were hosted by the present hostess.
When the difference between these two numbers was
positive, the invitee was a debtor to her hostess; when
it was negative, she was a creditor.
It is an open question whether the data will support
the market-specific or the general form of obligation.
Arguments can be made both ways. If market-specific
obligations such as parties are fungible with general
obligations such as child care and car pools, then the
latter may prove to be better predictors of purchase.
If, however, market-specific obligations are not fungible
with general obligations, then the former may
prove to be the better predictor.
Measures of UA. A measure for UA was created
from two indicators, one that describes the rate a respondent
consumed personal care products and another
that describes the respondent's preference for
the brand sold at the party. Prior to the party, respondents
were asked to describe the number of occasions
per week they used 12 types of personal care products
(e.g., lipstick, skin freshener, and so on). Respondents
were also asked to provide preference ratings for five
brands of each ofthe same 12 products (i.e., 1 = "totally
unacceptable," 5 = "acceptable," and 10 = "superb").
One of the five brands was the company
brand. A respondent's product consumption rates
were then multiplied by the difference between the respondent's
mean non-party brand preference and
company brand preference for the appropriate product.
Results were then summed across the 12 product
categories each respondent had rated.
Correlations between the indicators and two additional
measures supported the nomological validity
of the LO) indicators. Respondents also reported the
brand names ofthe 12 types of personal care products
they had used in a three-week period prior to completion
of the first questionnaire. The total number of
brands reported and the percentage of the total reported
that were company brands provided alternative
measures of respondents' personal care product
consumption rate and preference for the company
brand. The consumption rate indicator had a .53 correlation
with the total number of brands reported.
and the preference indicator had a .62 correlation
with the percentage of the total that were company
brands.
Income. Although not directly related to the theory
being tested, the ability to pay is an obvious source of
variation in the amount spent at a party. Invitees
from households with larger incomes are likely to
have more disposable income and therefore greater
freedom to make discretionary purchases for their
own benefit or for the benefit of their hostesses.
Household income was therefore included as a covariate
in the analysis.