Forced Random Search Decreasing Returns
Standard canonical search models often assume job search is undirected (random), predicting that lowering search costs across randomly encountered vacancies maintains baseline return rates. If platform vacancy ordering forces random search, the returns to marginal applications may differ fundamentally from directed search interventions.
Picture this
Think of browsing a menu where dishes are listed in random order rather than grouped by preference. If a waiter prompts diners to order whichever dish appears at the top of the page, diners buy items they enjoy less, leading to lower satisfaction per meal compared to when they choose their favorite options regardless of menu placement.
What the evidence says
Listing a vacancy first increased application probability by 0.440 percentage points (p < 0.001), but marginal applications yielded an interview rate of only 2.44% (p = 0.231 relative to 0%, p = 0.077 compared to 6.29% control return), demonstrating sharply decreasing returns compared to initiation-based search (5.87%).
- Who
- N = 9,255 jobseekers across 938,284 jobseeker-vacancy matches in Lahore, Pakistan (restricting to person-rounds with 2 or more matches).
- How
- RCT randomizing the listing order of matched vacancies on call/text notifications (in 20% of matching rounds) and estimating 2SLS returns to applications induced by top-of-list placement.
What to do
Avoid using random default sorting or top-of-list placement to drive user engagement on matching platforms, as forced random exposure induces low-yield actions with rapid return decay.
From the source
"Listing vacancies first produces more applications with decreasing, rather than constant, returns... the average interview probability for marginal applications submitted because the vacancy was listed first is 2.4%... sharply decreasing returns."
Why_Don’t_Jobseekers_Search_More_Barriers_and_Returns_to_Search.pdf