This paper studies how fixed hourly pay and piece‑rate pay (payment per task) affect worker effort and who chooses each option. Fixed wages make pay less risky for workers. But they can also lower effort (moral hazard) and change which workers choose fixed pay (adverse selection). The author ran a field experiment with online data‑entry workers to measure both forces and to estimate the welfare consequences.

In the experiment, workers on an online platform chose between a randomized fixed hourly wage and a piece rate. The data‑entry task involved typing short sentences for pay, with a small per‑correct‑sentence piece rate. The researcher controlled the offers and observed each worker’s choice and how much correct output they produced in a short timed task. The design aims to isolate how offers change both behavior and selection into contract types.

The results show evidence of both moral hazard and adverse selection. On average, accepting a fixed wage reduced worker productivity by about 8.74 percent relative to the sample mean. At the same time, raising the fixed wage attracted slightly more productive marginal acceptors: a 10 percent higher wage offer drew a marginal worker whose productivity was about 1.90 percent higher than the mean.

To interpret these patterns the paper uses a model that links moral hazard and adverse selection to what economists call marginal treatment effects (MTE). In plain terms, MTE looks at how small changes in offers change both who accepts and how those acceptors perform. The author then uses semi‑parametric MTE estimation to translate the measured changes into welfare numbers and to compute marginal values of public funds (MVPFs) for policies that might recover lost welfare.

One policy implication the paper reports is that taxing piece‑rate pay could reduce adverse selection into fixed wages and raise revenue while creating relatively low social costs. The author finds that tax revenue can be raised at social costs as low as $0.88 per dollar of revenue in this setting. That is, in the experimental market a modest tax on piece rates can partially correct the distortion created by asymmetric information and do so at less than a one‑to‑one efficiency loss.