Self-employment of older workers and the transition to retirement: using machine learning to uncover heterogeneity in work arrangements.

BACKGROUND AND OBJECTIVES: Self-employment increases with age, but heterogeneity in the self-employment arrangements of older workers and their transitions across these work arrangements is not well understood. This paper uses novel data to examine such heterogeneity before and after retirement.

RESEARCH DESIGN AND METHODS: We use machine-learning methods and internal narrative descriptions of industry and occupation in the 1994-2018 Health and Retirement Study to classify self-employment as (a) informal, (b) formal, and (c) business ownership. We then examine work and demographic characteristics associated with these different self-employment arrangements and how workers transition between these work arrangements across survey waves and at retirement.

RESULTS: Formally self-employed workers and business owners perform more abstract work and substantially less manual work and report more schooling, higher earnings, and better physical and mental health than informally self-employed workers, who are the most likely to transition out of the labor force by the next survey wave. Many self-employed workers transition into other self-employment arrangements across waves and at retirement and are less likely than employees to transition out of the labor force at retirement. Substantial shares of the formally self-employed, business owners, and employees pre-retirement transition into informal self-employment at retirement.

DISCUSSION AND IMPLICATIONS: Our findings reveal important differences in older workers' self-employment arrangements and transition patterns that were previously unobservable. Policymakers should be sensitive to this distinction, as policies targeting small business owners may not address the needs of other self-employed workers.