Business Networking and Social Cohesion: Experimental Evidence from Türkiye

This project uses a field experiment to evaluate a business matchmaking program in Gaziantep that
connects Turkish- and Syrian-owned SMEs. We investigate the extent of interethnic networking
frictions and whether reducing these frictions affects interethnic business interactions and inter-
group relations.

2025-09-15

2026-09-15

Business networks, Business Interactions, Business Performance, Social cohesion.

We study a randomized controlled trial of a business matchmaking program that connects Turkish
and Syrian SMEs in Gaziantep. Turkish firms are randomly assigned to one of four groups: (i)
an inter-ethnic treatment arm (Turkish–Syrian matches), (ii) an intra-ethnic treatment arm (Turk-
ish–Turkish matches), (iii) a passive pool of Turkish firms serving only as potential partners, and
(iv) a control group. Parallel groups are defined for Syrian firms, though with smaller sample sizes.
Our main focus is the Turkish sample.
All firms that are not in the passive pools (Chooser firms) complete a preference survey inwhich they specify the types of businesses they would like to meet. Matches are then generated using an LLM-based algorithm that ranks potential partners based on these preferences and a manual verification process. Each chooser firm is offered up to the five best matches selected from its randomly assigned pool of Turkish or Syrian firms. Chooser firms are first presented with the attributes of each potential partner without revealing identities and asked whether they are interested in meeting. After preferences are recorded, the identities (firm and owner names, making Syrian ownership clear) are revealed, and we ask whether the firms are already acquainted. Finally, chooser firms select which matches they would like to meet, and outreach officers contact the chosen firms to arrange meetings at the Gaziantep Chamber of Industry. Firms in the control group are also surveyed for their matchmaking preferences and their knowledge on the offered potential matches but are not called back for meetings. Meetings are facilitated by outreach officers, including a bilingual Turkish–Arabic officer who provides translation when needed. After each meeting, outreach officers immediately complete a structured survey that provides a detailed summary of the meeting. A follow-up phone survey is conducted with firms at two weeks after the meeting. These measure perceived match quality (trust, communication, alignment of practices, product quality), continued engagement (communication or transactions after the meeting), attitudes toward Syrian- owned businesses, and satisfaction with the service.

Finally, we plan to collect midline data approximately 3–4 months after the initial preference survey. The midline survey will include all firms, regardless of treatment assignment or meeting
participation, to assess business activity and intergroup perceptions. In an updated version of this
registry, we will pre-register our decisions regarding the outcomes to be measured and any potential
adjustments to the identification strategy shortly before the midline data collection begins.

Not available

Randomization is done by a computer in the office using a random number generator in Stata.

Firms

Yes

3,011 firms: 2,315 Turkish-owned and 696 Syrian-owned businesses.

We observe pre-meeting outcomes at the firm×match level. Since each firm is offered the opportunity to evaluate up to five potential matches, we expect the number of observations to increase. proportionally. For the meeting and post-meeting outcomes, the data are at the firm level, and thus the number of observations should be equivalent to the number of clusters reported in the previous section.

Treatment 1: Intra-ethnic Chooser (Turkish firms): 722
Treatment 2: Inter-ethnic Chooser (Turkish firms): 722
Treatment 3: Intra-ethnic Chosen (Turkish firms): 550
Treatment 4: Inter-ethnic Chosen (Syrian firms): 550
Treatment 5: Control (both ethnicities): 467