Two-Step Cluster Randomization Design

In Simple Words

Imagine playing a game of musical chairs in a room where shoes with springs are given to half the players. Those with spring shoes grab chairs faster than their barefoot room peers, but because the total number of chairs stays identical, the barefoot players end up without seats. To discover whether the shoes actually create extra chairs or merely swap who gets seated, one must compare a room where some people received spring shoes against a completely separate room where nobody received spring shoes.

Research question

Standard individual-level randomized controlled trials fail to detect market-wide spillover and displacement effects because untreated control units interact with treated units in the same local job market. A two-step cluster randomization design is required to isolate individual treatment gains from market-wide displacement effects by varying treatment saturation across geographical areas.

Participants

Nearly 30,000 young, college-educated human job seekers in their mid-twenties possessing at least a two-year degree and unemployed for at least six months across 235 communities in ten regions of France.

Study Design

Randomized controlled trial using a two-step cluster randomization design conducted from 2007 to 2010, where local communities were first randomly assigned a treatment saturation level (0%, 25%, 50%, 75%, or 100%) and eligible individuals within treated communities were then randomly assigned to treatment or comparison groups.

Results

While individuals offered counseling were 2.5 percentage points more likely to find long-term employment after eight months than untreated peers in the same communities, untreated peers in treated communities were 2.1 percentage points less likely to find jobs than job seekers in pure comparison communities where 0% were treated.

Takeaway

Design active labor market evaluations using a two-step cluster randomization framework that varies treatment saturation across distinct local market clusters to measure spillover effects.

Read alongside the research

Plain-English glossary

Original explanations, informed by the references below. Examples are illustrations, not data from the studies above. The research text stays unchanged.

Percentage points

The difference between two percentages.

ExampleA rise from 40% to 50% is 10 percentage points. Relative to the starting value, that is a 25% increase.

Office for National Statistics · Percentages ↗Read more about percentage points →

Randomized controlled trial

A study that uses chance, like drawing lots, to decide who receives each option being compared. Researchers compare the groups’ outcomes to estimate the effect of those options.

ExampleA lottery assigns volunteers to a training programme or a comparison group. Researchers then compare their outcomes.

GET-IT glossary · Randomized study ↗Read more about randomized controlled trial →

Related findings

• Advance Notification Anticipation Effect →Labour & employment

  Imagine being told weeks in advance that a strict inspector will audit your progress on a specific date. To avoid the stress and potential penalties of the audit, you work harder and accept acceptable options earlier to resolve the situation before the audit date arrives.

• Auxiliary Program Decoupling →Labour & employment

  Imagine testing whether signing a fitness pledge helps people exercise. If participants who sign the pledge also receive free personal training sessions, it is impossible to know whether the pledge itself worked or the extra coaching produced the results.

• Displacement Externalities in Active Labor Market Programs →Labour & employment

  Imagine a crowded musical chairs game where some players are suddenly given fast running shoes. While those runners grab available chairs much faster, they bump into and push away the barefoot players, making it significantly harder for the non-runner players to find a seat.

• Double-Sided Randomization Design →Labour & employment

  Imagine a musical chairs game where organizers want to know if giving subsidized tickets to new players adds total active players or merely pushes out un-subsidized participants. By flipping a coin to select both which players receive subsidized tickets and which chairs host subsidized players, organizers can directly observe whether new total chairs are created or existing participants are displaced.