Web Appendix A: List of Journals

Web Appendix A: List of Journals

Tables WA1, WA2 and WA3 present the list of all journals in which job candidates were observed to publish up to one year after being hired. These journals span various disciplines: management, marketing, economics, statistics, and others. The total number of publications in each of these outlets is also displayed.

TABLE WA1. Job Candidate Publications in Top Marketing Journals

Job Markets 1997 – 2005

TABLE WA2. Job Candidate Publications in Other Marketing Journals

Job Markets 1997 – 2005

TABLE WA3. Job Candidate Publications in Non-marketing Journals

Job Markets 1997 – 2005

(Continues on next page)

TABLE WA3. Job Candidate Publications in Non-marketing Journals (cont.)

Job Markets 1997 – 2005

Web Appendix B: Identification in the Fox Model

Identification in the Fox Model is contingent on the type of variables included in the specification of the local production maximization condition, as shown in Equations 3 and 4. Interactions between continuous variables are straightforward, as these involve no identification problem. Interactions between discrete-discrete and discrete-continuous variables, however, pose some identification restrictions. These two cases are discussed in what follows.

Case 1. Discrete-discrete interactions

The intuition behind a set of discrete-discrete interaction is that agents belong to different categories on each side of the market. These are then interacted to form bins or match types. Each observed match, then, can be said to belong to one of these bins. The matching value of each bin can then be estimated.

To illustrate, consider the situation where equilibrium matches and are observed. Suppose candidates and departments can be described by their ranking status (see Table 1). Denote the ranked status of a candidate as variable and the ranked status of a department as variable , where . Also let variable indicate the non-ranked status of a candidate and variable indicate the non-ranked status of a department. These four variables are dummies that take the value of "1" if a certain agent is of a given type and 0 otherwise. In addition, describe candidates using their universities' financial endowment , and describe departments in a similar fashion using their endowment . The parameters that could be potentially estimated from the structural model are shown in Table WB1.

TABLE WB1. Interaction bins and parameters, discrete-discrete case

Identification can be explored by examining the local production maximization conditions generated from all possible matches between different types of agents. A different parameter is associated with each different type of match shown in Table WB1. The final resulting local production maximization conditions from these types of matches are shown in Table WB2.

TABLE WB2. Local production maximization conditions, discrete-discrete case.

Shown in Table WB2 are the conditions characterizing an observed match of job candidate C1and hiring department D1, as well as the conditions for an observed match of job candidate C2 to hiring department D2.Each row in the table, therefore, represents two observed matches whose members may be similar or different. Each agent is characterized using one covariate (ranked status) and one continuous characteristic (university financial endowment). A ranked candidate (department) is denoted by RC (RD). and a non-ranked candidate (department) is denoted by NRC (NRD). The candidate’s school's financial endowment is denoted by , and the hiring department’s school’s financial endowment is denoted by . To simplify our analysis, the notations and denote and , respectively. Finally, the ω terms represent the matching value resulting from the interaction between ranked and non-ranked status corresponding to Table WB1.

The combinations of matches above are used to assess the structural condition statedin Equation 1and specified in Equation 4in the manuscript, considering only the rank status of the candidate and hiring department and their respective endowments. Thus, for two candidates and 2 departments, where each could be ranked or unranked, there are 16 different possible configurations, corresponding to the 16 rows of Table WB2.

The third column of Table WB2 represents the resulting structural condition implied by each of the 16 possible match combinations. Only certain types of match pairs are informative for estimation. In this case, only four such pairs can inform the model on the value of , , and . Importantly, the results in Table WB2 show that only one of these four parameter can be identified. This is why , and are set to zero in the estimation results presented in this paper.

Case 2. Discrete-continuous interactions

The intuition behind a discrete-continuous interaction is that agents on one side of the market belong to different categories, but agents on the other side are described by a set of continuous variables.Returning to the example and notation used in Case 1, now let candidates be described by their fields with three variables, , and , where . Similarly, let departments be described by the number of their faculty trained in a certain field, denoted as, and . Note that the latter variables are continuous and exhibit no linear dependency. The possible interactions that result from the discrete-continuous case are no longer bins. These are presented in Table WB3.

TABLE WB3. Interactions and parameters, discrete-continuous case

Identification can be explored by examining the local production maximization condition that follows from all the cases of different competing candidates only. That is, only the side of the market with discrete interactions needs to vary, since departments no longer belong to different categories or bins. This is shown in Table WB4.

TABLE WB4. Local production maximization conditions, discrete-continuous case

Table WB4 shows the conditions characterizing an observed match of job candidate C1 and hiring department D1, as well as the conditions for an observed match of job candidate C2 to hiring department D2, similar to Table WB2. Each row in the table, again, represents two observed matches whose members may be similar or different. Each agent is characterized using one covariate (field of research) and one continuous characteristic (university financial endowment). Consequently, candidates are characterized by a discrete variable (which field of research they belong to) and hiring departments are characterized using three continuous variables (the number of CB, Modeling and Strategy scholars in the department, respectively). The values and are equivalent to and , respectively, following the notation of Table WB3. Financial endowment notations are identical to Table WB2. The ω terms represent the matching value resulting from the interactions corresponding to Table WB3.

The combinations of matches above are also used to assess the structural condition stated in Equation 1and specified in Equation 4 in the manuscript as well. These consider field of research and endowments. For two candidates and two departments, there are 9 different possible configurations: as stated above, only candidate swaps matter, because the hiring department side’s covariate is continuous.

Two sets of results are shown. The first set shows the case where no parameter is restricted. Closer inspection reveals the system is not identified. The second set shows the case where the parameters corresponding to Strategy candidates are set to zero.

The third column of Table WB2 represents the resulting structural condition implied by each of the 16 possible match combinations. Only certain types of match pairs are informative for estimation. In this case, only four such pairs can inform the model on the value of , , and . Importantly, the results in Table WB2 show that only one of these four parameters can be identified. This is why , and are set to zero in the estimation results presented in this paper. Once the parameter restriction is in place, then the remaining parameters can be successfully identified.

Web Appendix C:Data correlations

The following table presents the variables included in the correlations table

TABLE WC1. Variables included in the Correlations table

Job Markets 1997 – 2005

Next, Table WC2 presents the correlations among the variables used in our dataset. These correlations comprise data from the years 1997-2005. Variable numbers correspond to those shown in Table WC1.

TABLE WC2. Correlations - Marketing job market dataset

Job Markets 1997 – 2005

Web Appendix D:Top-Ranked Status – Categorized as Top 20

In the paper, top-ranked status is assigned to a candidate if he or she graduated from a business school that was listed at least once at any of the top 30 spots in the Financial Times (FT)or the UTD rankings during the years 2000 – 2005. Here, we assign top-ranked status if a candidate graduated from a top 20 school (instead of top 30). This serves as a robustness check of our models.

Table WD1. Model 1—The Base Model

Candidate human brand cues displayed in rows; department brand cues displayed in columns.

N=677. Significant estimates indicated with * (p<0.05).

Table WD2. Model 2 – The Matthew Effect

Candidate human brand cues displayed in rows; department brand cues displayed in columns.