This site is a work in progress and has not been widely shared. Content may contain errors. Feedback is welcome.
This site is undergoing review. Some annotations were human-generated, some AI-generated — all are being verified.
Back to papers

Monopsony in Online Labor Markets

Dube, Jacobs, Naidu, Suri

2018NBER Working Paper Series48 citations
Experimental evidenceCausalTheoretical model
Platforms / gig economy
View RepositoryDOI: 10.3386/w24416
Abstract

On-demand labor platforms make up a large part of the "gig economy."We quantify the extent of monopsony power in one of the largest on-demand labor platforms, Amazon Mechanical Turk (MTurk), by measuring the elasticity of labor supply facing the requester (employer) using both observational and experimental variation in wages.We isolate plausibly exogenous variation in rewards using a double-machine-learning estimator applied to a large dataset of scraped MTurk tasks.We also re-analyze data from 5 MTurk experiments that randomized payments to obtain corresponding experimental estimates.Both approaches yield uniformly low labor supply elasticities, around 0.1, with little heterogeneity.

Summary

Dube, Jacobs, Naidu, and Suri use randomized wage experiments and scraped observational data from Amazon Mechanical Turk (2014-2017) combined with double machine learning to estimate labor supply elasticities facing individual requesters and measure monopsony power in online labor markets

Main Finding

Requesters on MTurk face highly inelastic labor supply with elasticity around 0.1, indicating substantial monopsony power; a 10% wage increase reduces time to fill jobs by only 1%, allowing requesters to pay wages well below marginal product

Primary Datasets

Amazon MTurk data

Secondary Datasets

Worker surveys

Key Methods
Field experiments with randomized wage variation ('honeypot' design), observational data from MTurk scraped 2014-2017, double machine learning to control for task characteristics, random forest regression
Sample Period
2015-2018
Geographic Coverage
International
Sample Size
258,352 HIT groups (2014-2016 scrape); 292,746 HIT groups (2016-2017 scrape); 93,775 HIT groups (2017 scrape); multiple experimental studies surveyed and replicated
Level of Analysis
Individual, Firm
Occupation Classification
N/A
Industry Classification
N/A
Replication Package
Yes
Notes
Labor supply elasticity; Non-US paper; included for cross-national comparison [Claude classification]: Uses 'honeypot' experimental design where workers see randomized wages for identical tasks. Double machine learning with random forests to control for task characteristics in observational data. Combines experimental recruitment studies with retention experiments. Text features extracted using n-grams, LDA topic models, and Doc2Vec embeddings. Labor supply elasticity around 0.1 indicates monopsony markdown of approximately 10:1 (wage to marginal product ratio) [Claude classification]: Uses 'honeypot' experimental design where workers see randomized wages for identical tasks. Double machine learning with random forests to control for task characteristics in observational data. Combines experimental recruitment studies with retention experiments. Text features extracted using n-grams, LDA topic models, and Doc2Vec embeddings. Labor supply elasticity around 0.1 indicates monopsony markdown of approximately 10:1 (wage to marginal product ratio) [Claude classification]: Uses 'honeypot' experimental design where workers see randomized wages for identical tasks. Double machine learning with random forests to control for task characteristics in observational data. Combines experimental recruitment studies with retention experiments. Text features extracted using n-grams, LDA topic models, and Doc2Vec embeddings. Labor supply elasticity around 0.1 indicates monopsony markdown of approximately 10:1 (wage to marginal product ratio) [Claude classification]: Uses 'honeypot' experimental design where workers see randomized wages for identical tasks. Double machine learning with random forests to control for task characteristics in observational data. Combines experimental recruitment studies with retention experiments. Text features extracted using n-grams, LDA topic models, and Doc2Vec embeddings. Labor supply elasticity around 0.1 indicates monopsony markdown of approximately 10:1 (wage to marginal product ratio) [Claude classification]: Uses 'honeypot' experimental design where workers see randomized wages for identical tasks. Double machine learning with random forests to control for task characteristics in observational data. Combines experimental recruitment studies with retention experiments. Text features extracted using n-grams, LDA topic models, and Doc2Vec embeddings. Labor supply elasticity around 0.1 indicates monopsony markdown of approximately 10:1 (wage to marginal product ratio) [Claude classification]: Uses 'honeypot' experimental design where workers see randomized wages for identical tasks. Double machine learning with random forests to control for task characteristics in observational data. Combines experimental recruitment studies with retention experiments. Text features extracted using n-grams, LDA topic models, and Doc2Vec embeddings. Labor supply elasticity around 0.1 indicates monopsony markdown of approximately 10:1 (wage to marginal product ratio) [Claude classification]: Uses 'honeypot' experimental design where workers see randomized wages for identical tasks. Double machine learning with random forests to control for task characteristics in observational data. Combines experimental recruitment studies with retention experiments. Text features extracted using n-grams, LDA topic models, and Doc2Vec embeddings. Labor supply elasticity around 0.1 indicates monopsony markdown of approximately 10:1 (wage to marginal product ratio) [Claude classification]: Uses 'honeypot' experimental design where workers see randomized wages for identical tasks. Double machine learning with random forests to control for task characteristics in observational data. Combines experimental recruitment studies with retention experiments. Text features extracted using n-grams, LDA topic models, and Doc2Vec embeddings. Labor supply elasticity around 0.1 indicates monopsony markdown of approximately 10:1 (wage to marginal product ratio) [Claude classification]: Uses 'honeypot' experimental design where workers see randomized wages for identical tasks. Double machine learning with random forests to control for task characteristics in observational data. Combines experimental recruitment studies with retention experiments. Text features extracted using n-grams, LDA topic models, and Doc2Vec embeddings. Labor supply elasticity around 0.1 indicates monopsony markdown of approximately 10:1 (wage to marginal product ratio) [Claude classification]: Uses 'honeypot' experimental design where workers see randomized wages for identical tasks. Double machine learning with random forests to control for task characteristics in observational data. Combines experimental recruitment studies with retention experiments. Text features extracted using n-grams, LDA topic models, and Doc2Vec embeddings. Labor supply elasticity around 0.1 indicates monopsony markdown of approximately 10:1 (wage to marginal product ratio) [Claude classification]: Uses 'honeypot' experimental design where workers see randomized wages for identical tasks. Double machine learning with random forests to control for task characteristics in observational data. Combines experimental recruitment studies with retention experiments. Text features extracted using n-grams, LDA topic models, and Doc2Vec embeddings. Labor supply elasticity around 0.1 indicates monopsony markdown of approximately 10:1 (wage to marginal product ratio)