Demand for Automated Wage Adjustment

a) Offer of automated wage setting tools, with randomized features (described below).
b) Randomization of signal order (fill/unfill and large/small pool of potential applicants)

2025-04-06

2025-10-01

Demand for automation tools (described below)

This will be a binary variable equal to the share of firms indicating interest in each automated wage adjustment tool.

Perception of tools’ impact on filling position, and cost of filling position

Belief updating (conditional on a signal: filling a job on-time, or not) about candidate pool size, relative market wages, and response of fill-rate to wage adjustments.

Perceived cost of an unfilled job

We ask about beliefs over wage bill cost and fill effects for each tool. Independent of the tools, we elicit beliefs before and after jobs go unfilled and when jobs go unfilled in a market with few or many job seekers. The order of fill/unfill questions and the thick/thin market questions is randomized and cross-randomized between the two groups.

We study firms' willingness to adopt particular automated wage adjustment tools we create, and under what conditions. Participating firms are introduced to a set of wage-setting tools and asked to indicate their interest in using each one, under the clear and truthful premise that the staffing firm eliciting this information will use responses to inform the design of their automated wage-setting tools and roll-out of these tools. In other words, respondents are informed that their expressed preferences may influence their experience on the staffing platform they use, creating an incentive-compatible setting.

The main design is a within subject comparison, contrasting the search behavior of a firm when a specific automated wage setting tool is available with their search under current conditions (without wage setting automation) or with an alternative wage setting tool. The tools offered vary along dimensions such as the increment and timing of automated adjustment and the conditions that trigger the adjustment. The order in which the tools appear is randomized across subjects. The available tools are:

AutoIncrease – A baseline automated tool that increases advertised pay by 10% for unfilled positions as the intended start date approaches.

AutoBonus – Similar to AutoPay but frames the increase as a hiring bonus rather than a wage increase.

CompAssist – Automatically raises pay by 10%, with the full cost covered by the staffing platform.

MarketPilot – Triggers a 10% wage increase when the position remains unfilled close to the start date, but only if, in addition, at least 50 qualified workers have seen the position but accepted jobs elsewhere.

We compare demand for market tools, and implied rate of wage adjustments and fill rates, with the no-automation-tool status quo.

To explore mechanisms we test for incomplete information: Whether subjects update about candidate pool size, relative market wages, and response of fill-rate to wage adjustments as a function of whether they fill a job on-time, or not (two hypothetical scenarios that we randomize)

We also elicit the cost of an unfilled job by asking about lost revenue, and willingness to pay a premium to fill by the start time.

Randomization will occur via Qualtrics.

We anticipate there will be around 800-1,000 firms in the experiment (perhaps around 200-300 firms on the focal platform and around 600 on Prolific). This will depend on the share of firms that are willing to participate.

Yes

Around 800-1,000 hiring managers.

Around 800-1,000 hiring managers.

Each hiring managers will be asked about 4 types of wage adjustment. All options are shown simultaneously and we randomize the order of the list.