Did Algorithms Replace the Hometown Network? Food-Delivery Platform Entry, Job-Search Channels, and Migrant Workers’ Incomes in China

Abstract Does algorithmic matching replace informal job-search networks? We exploit the staggered, month-precise entry of food-delivery platforms into Chinese counties, using the China Migrants Dynamic Survey (2011 to 2018, about 1.29 million respondents). Platform entry did not measurably alter channel composition: power analysis rules out effects beyond about one percentage point for the internet channel and four points for the tie channel, while smaller declines cannot be excluded. The shock was economically real: migrant incomes rose 4 to 6 percent in an imputation event study, surviving the exclusion of rider-adjacent and restaurant workers, with larger gains for high-education migrants. Whether the return to social capital declined cannot be tested with the survey proxy: the interaction estimate fails pre-trend, in-time placebo, and randomization-inference diagnostics. Algorithms raised incomes without restructuring search. JEL codes : J61, J64, O33, D85