party was 6.40, whereas the mean preference for the
non-party brands was 5.75, indicating that, on average,
invitees preferred the party brand slightly more
than the other brands.
Second, there were many hostess-invitee and some
demonstrator-invitee relations strong enough to generate
UE. Some 59 percent of the IH relations were
classified as friends, close friends, or relatives,
whereas only 5 percent ofthe ID relations were similarly
classified. This is as one might expect, since IH
relations are part of the hostesses' existing personal
networks, whereas ID relations are probably formed
in the course ofthe home party.
Third, the data indicate that many invitees had
market-specific obligations with their hostesses. Approximately
half of the invitees had hosted at least
one party that the present hostess had attended. Of
those who had hosted previously, 62 percent owed the
present hostess one or more parties, 32 percent were
even, and only 6 percent were net creditors. Far fewer
invitees had general obligations. The data indicate
that 80 percent ofthe invitees considered themselves
"about even," 10 percent were debtors, and 10 percent
were creditors.
Finally, there are indications that some portion of
the sample was able to achieve satisfactory levels of
UT, as evidenced by the purchases completed. The
data indicate that 97 percent of the attending guests
made a purchase, buying an average of four items for
a total average expenditure of $30.70. Purchases
ranged from a low of $ 1 to a high of $ 124, with a standard
deviation equal to $24.30.
PROBIT Results
Coefficients for the first step in Heckman's procedure,
the PROBIT analysis, appear in Table 1.* The
Table 1 contains only a main efTects model. Interaction terms
among and between the U^ and the UE variables were tested, and
UA variable bears a positive sign, indicating that the
UA offered by the party products increased the likelihood
an invitee entered the market and purchased.
Note that the income covariate is also related to the
likelihood of purchase, as one might expect.
The weak form hypotheses, which describe the influence
of the UE variables on the likelihood of purchases,
were both supported. The strength of both IH
and ID relations exerts a significant and positive influence
on the likelihood of purchase. The magnitude
and significance ofthe IH and ID coefficients are virtually
identical. Table 1 also shows that market-specific
obligation exerts significant, positive effects on
purchase likelihood. According to the coefficients, a
unit increase in the outstanding obligation of a single
party has a much larger effect on purchasing likelihood
than a unit increase in IH tie strength. The general
obligation variable failed to have a significant impact
on the likelihood of purchase.
The results imply that the UE variables affect market
entry in the manner hypothesized by the weak
form hypotheses, even when the influences of traditional
market factors (such as income and UA) are
controlled. As predicted, buyers with strong social
ties to sellers and market makers, such as hostesses,
are more likely to enter the market. Further, in this
study, buyers who were obligated to the market maker
were more likely to enter, but the type of obligation
appears to be peculiar to the home party market. This
result, however, may be due to low variance in the
general obligation measure.
Bias-Corrected Regression Results
Coefficients for the second step of Heckman's procedure,
the bias-corrected regression, are included in
all were found to be insignificant. It is quite possible, however, that
the assumption of a strictly additive model may need to be relaxed
in other market situations.