In this article:

The present paper has considered an important empirical feature of the labour market, namely that among those in work, self-employment rates are higher for those with work-limiting disabilities than for the non-disabled. This was argued potentially to reflect the opportunities self-employment might afford for the former to accommodate their disability by choosing times, hours and locations of work.

Self-employment was modelled using a bivariate probit model allowing for the possibility of selection effects into employment for each of three disability groups by gender along the lines suggested by DeLeire (2001), namely the non-disabled, the work-limited disabled and the non-work-limited disabled. In general results are highly intuitive, and predicted employment and conditional self-employment probabilities mirror closely those observed in the data.

The DeLeire technique allows for the separation of the effect of accommodation from the influence of discrimination on conditional self-employment probabilities. After controlling for characteristics, the non-work limited disabled are less likely to be in self-employment than the non-disabled, which suggests consumer discrimination may be important. However, for men we find strong evidence of the work-limiting nature of disability being an important positive influence on self-employment, consistent with the greater ability to accommodate a disability in self-employment.

You can read the complete article here.