Experimental Design Details
We will use fliers to recruit both firms and individuals interested in participating in the internship program. One flier targets firms in the finance industry seeking interns, asking them to specify the position’s location at the city level, one primary required skill (e.g., data analysis), and the intern’s academic field (e.g., economics). A separate flier targets students and recent graduates in business, finance, economics, or related fields, inviting them to provide basic personal information, academic field, educational background, work experience, one primary skill (e.g., data analysis), and preferred internship location at the city level. The subsequent stages of the experimental design are as follows.
After recruitment, we will have two groups of participants: (1) individuals interested in internships (referred to as “interns”) and (2) firms seeking interns. Suppose that (M) interns and (N) firms express interest in participating. Among the (M) interns, a fraction (\alpha) are female, so there are (\alpha M) female and ((1 - \alpha) M) male interns.
To match firms and interns, we use three key pieces of information for each intern: their preferred city, primary skill (e.g., data analysis), and academic field (e.g., economics). Similarly, for each firm, we record the city where the internship is offered, the primary skill required, and the academic field relevant to the position. These three characteristics—city, skill, and field—jointly define a “cell.” For example, one cell might represent internships in *Tehran–Data Analysis–Economics*.
We then group interns and firms into cells based on these three characteristics. Within each cell, we identify all interns and firms whose information exactly matches that combination. For example, suppose in one cell there are (X) interns and (Y) firms.
Within each cell, we randomize the order of interns and firms separately. That is, we assign a random rank to each intern from 1 to (X) and a random rank to each firm from 1 to (Y). Randomization is conducted in the office using a computer.
Next, we create one-to-one matches between interns and firms in that cell. Matching is done sequentially based on the random rankings: the highest-ranked firm is matched with the highest-ranked intern, the second-ranked firm with the second-ranked intern, and so on. The number of pairs that can be created in each cell depends on the smaller of the two groups. If there are more interns than firms ((X > Y)), only the top (Y) interns are matched, and the remaining (X - Y) interns remain unmatched. Conversely, if there are more firms than interns ((Y > X)), only the top (X) firms are matched, and the remaining (Y - X) firms remain unmatched.
This process is repeated independently for every city–skill–field combination. The result is a set of one-to-one matches between firms and interns that are perfectly aligned in terms of location, skill, and field preferences. Each participant (firm or intern) is used at most once, and no cross-cell reallocations are made.
Lastly, we will share the matching results with both sides. Firms will receive the following information about their matched intern to make a decision: academic field, educational background, work experience, one primary skill (e.g., data analysis), and preferred internship location at the city level. No identifying information, gender, or age will be provided. Interns will be informed of their matched firm, including the firm’s name and the exact location of the internship position. After both sides accept the match, each will receive the contact information of the other party and may follow up to begin the internship.
Both firms and applicants retain the right to accept or decline the proposed match and may discontinue the internship at any time and for any reason.