Friends in High Places: The Effects of Social Networks on Discrimination in Salary Negotiations

Administrative Science Quarterly,  June, 2000  by Marc-David L. Seidel,  Jeffrey T. Polzer,  Katherine J. Stewart

• E-mail
• Print
• Link

Supplemental Analyses

The effect of the control variable other offers on negotiated percentage salary increase was unexpectedly negative, although it did not reach significance in the fully specified model. Because alternatives to a negotiated agreement are frequently cited as an important factor in negotiations, we used OLS regression to determine what factors influenced the number of other offers held by an applicant. As reported in table 4, quality rating had a significant but small positive effect, as we would expect. Race was a much larger predictor. Using White as the excluded reference group, both Blacks and Hispanics had significantly more offers than Whites, perhaps because qualified minority applicants are highly sought after and thus receive many job offers. Managerial comments on minority recruiting support this view (Angwin and Castaneda, 1998). Although we would expect applicants with more alternatives to have greater bargaining power in the negotiation, the negative coefficient for other offers in table 2 indicates that the other offers held by Blacks and Hispanics did not result in higher salary gains through the negotiation process. Apparently, Blacks and Hispanics were unable to translate their higher number of other offers into superior negotiated outcomes. While highly qualified minority applicants may be highly sought after, it is not clear that this effectively translates into negotiated outcomes. We can only speculate as to why this is. One possibility is that minorities' other offers are not good offers and are thus not good alternatives to a negotiated agreement. Another possibility is that those applicants not referred by friends do not realize the extent to which other offers provide additional bargaining power or do not incorporate this additional power into their negotiation in effective ways. This is consistent with the network argument.

Although our theory focuses on negotiated increases, we also conducted an analysis of the effects of our independent variables on initial offers. We did not include other offers as a control variable in these equations because this information was not collected until applicants had accepted a job, and other offers could have been received after the initial offer was made. As shown in table 5, age, education, and OPI all had positive effects on the initial salary offer, as we would expect. Using White as a reference group, there were no significant effects for minority group membership on initial offer in model 2, nor did having a social tie to the organization affect initial offers when entered alone in model 3. When friend and racial categories were both included in the model, however, the effect for Black applicants became significant, indicating that Black applicants received significantly lower initial offers than White applicants.

Because the results of the logistic regressions predicting method of referral (table 3) revealed that race had differential effects on all methods of referral, we tested the effects of these methods of referral on initial offer (table 5, models 5 and 6) and on negotiated percentage salary increase (table 6). In each set of regressions, we used friend as the excluded reference group to test the dummy variables for each method of referral. For initial offer, applicants who were referred to the company by a headhunter received significantly higher initial offers than those referred by a friend, while applicants who gained access to their job through a campus interview received significantly lower initial offers. When both race and method of referral were included in the equation in model 6 of table 5, the positive effect for using a headhunter remained significant. The significant negative effect for Black applicants found in model 4 of table 5 was no longer significant in model 6 when all methods of referral we re included in the equation. This shift in significance can be explained by the results in table 3, showing that Black applicants were significantly more likely than White applicants to be referred to the company through a headhunter. Therefore, in model 6, the inclusion of headhunter in the equation accounted for some of the variance that was attributed to the variable for