It has been argued that declining housing affordability in Australia’s major cities has led to the exclusion of many low- and moderate-income residents from high employment, inner-city regions. If there is an increasing spatial mismatch between housing and employment, moderately paid workers, essential to the efficient functioning of the urban economy, may face problems in accessing and retaining employment. However to date there has been a lack of empirical analysis of the spatial dimensions of housing and employment (and the commuting such divisions necessitate) broken down by occupation. Using the aggregate 2001 Census Journey to Work data by Statistical Local Area (SLA), we apply a spatial aggregation algorithm to develop largely self-contained commuting areas in Sydney, Melbourne and Brisbane. We establish that these areas are also relatively self-contained with respect to commuting flows by occupation. We employ linear programming techniques to determine the spatial patterns of commuting by occupation within these metropolitan commuting areas which minimize the corresponding average distances commuted. The results reveal some variation in commuting patterns across occupations but little evidence of longer commutes for the low-skilled. The results highlight the need to separate out the ‘volitional’ or ‘excess’ component from the overall commute, particularly if relying on commuting data to make inferences about how considerations of housing affordability impact on the locational decisions of lower-income workers within metropolitan areas.