From the implementation of new and emerging technologies including artificial intelligence (AI) through to occupational shifts with reduced employment in traditional manufacturing and even some service jobs, Australian workers risk facing displacement from a range of sources. These risks can reduce the wages and job mobility of Australian workers, increase the risk of unemployment particularly among those in traditional routine task jobs, and potentially slow productivity and output growth. One possible avenue for supporting workers during this instability while improving their productivity is to improve their human capital; however, there has been limited research on how specific forms of human capital benefit Australian workers and production in modern technological environments. This thesis investigates soft skills as a form of human capital relevant in modern economies, due to how these skills simultaneously enable workers to better work in offices affected by these transitions and as these skills are difficult to automate. However, there is a knowledge gap on the strength of the benefit of soft skills, particularly for the Australian economy. As such, this thesis investigates whether these soft skills have previously benefited Australian workers and the broader economy via analyses from multiple perspectives on the Australian labour market and economy.

The first empirical chapter in this thesis investigates whether worker job mobility is associated with higher soft skills, as these skills are utilised across a range of occupations and thus should enable workers to reallocate to where they are most valued. This chapter utilises the Household, Income and Labour Dynamics of Australia (HILDA) data to investigate worker job mobility and how this is associated with a selection of soft skill proxies including time management, social capital, and low task repetitiveness in the job. A Cox survival analysis is chosen to test if soft skills are associated with a reduced duration between job-to-job transitions. This modelling finds none of the soft skill measures have a significant association with the duration between job transitions. By comparison, endowments of overall skills have a mixed relationship: education is primarily relevant through the positive relationship between vocational education and duration, although some specifications suggest school education is also positively associated; meanwhile, having a job with a higher ability to decide what to do in it is positively associated with lower mobility. When further testing how this job mobility correlates with wages in a linear random effects model with Mundlak corrections and interactions between job mobility and skills, this research finds job mobility has no significant association with wages, excluding one interaction term with university education that has a negative association with wages. These results find soft skills have no broader relationship with wages, with only low task repetitiveness having a significant and positive association with wages.

The second empirical chapter in this thesis investigates how soft skills are associated with unemployment risk, also using the HILDA data and replacing communication skills with social capital as a form of soft skill development. This chapter focuses on the Global Financial Crisis (GFC) as being associated with further reducing the risk of unemployment for workers with higher soft skills, based on skill-biased technological change (SBTC) or routine-biased technological change (RBTC). As businesses may use alternative methods to control costs or fill necessary employment gaps during a negative economic shock, additional specifications are tested that include underemployment and overskilling as intermediate levels of labour match between well employed and unemployed. This chapter also utilises the HILDA data to determine the risk of unemployment in nonlinear random effects logit regression models with Mundlak corrections and year effect interactions to test for year-by-year changes in these skills. The results find social capital has a significant association with reducing all negative outcomes, while the remaining soft skill proxies demonstrate insignificant results for all specifications bar overskilling. In this overskilling model, low task repetitiveness is associated with a lower risk of overskilling; however, this appears to stem from job attributes instead of underlying skills. Similarly, time management has a significant and positive association with unemployment risk, appearing to stem from the underlying construct rather than time management itself. Further, the overskilling specification has significant year interactions for all skill measures from 2009 to 2015, potentially suggesting SBTC and RBTC occurred and had lingering effects. Of these interactions, both time management and social capital have negative associations with overskilling risk, reinforcing the potential effects of the GFC on skill usage for Australian workers. Overall, this is some evidence in favour of soft skills having an effect on workers, albeit constrained by the significance primarily within the overskilling model and the overall positive association between time management and overskilling.

The third empirical chapter investigates the industry perspective to determine whether soft skills are associated with increased productivity and output. To investigate SBTC and RBTC in particular, this chapter includes information communication technology (ICT) as both an individual variable and with a moderating interaction with overall soft skill level in some specifications. Seven skills formed the basis for the soft skill component: initiative and innovation, learning ability, oral communication, planning and organising, problem solving, teamwork, and writing as a measure of written communication. Models were tested using two specifications of soft skills: an overall soft skill measure generated using confirmatory factor analysis (CFA) for investigating the overall relationship and for the moderating relationship with ICT, and the individual measures for more detailed investigation of individual skill relationships. These specifications were tested for three measures of industry capability: multifactor productivity (MFP), overall value-added output, and per worker value-added output. Based on the Australian Skills Classification (ASC) and Australian Bureau of Statistics (ABS) data utilised, a short-run fixed effects model estimated using ordinary least squares (OLS) was used to test these specifications. This research finds overall soft skills and their moderating interaction with ICT are insignificant. When broken down into individual soft skills, oral communication is associated with higher industry value added and written communication is associated with lower industry value added, with no other soft skills of significance. This suggests soft skills are of limited value in the short-run. Further, oral and written communication are also only significant in the overall value added model, suggesting potential industry-wide effects of deep labour pools are more relevant than the aggregate skill level of the workers.

These results suggest the effect of soft skills on the Australian economy varies depending on the type of skill measure and the economic outcome governments to be improved, with the conclusion detailing policy recommendations and theoretical contributions. For the practical applications, soft skills have limited effect at the macroeconomic level beyond overskilling and to a lesser degree unemployment risk, thus training should depend on what is required for the job or education program. However, broader communication skills and social capital did demonstrate significance across multiple specifications, reinforcing the relevance of these particular skills. The significance of social capital highlights the importance of developing soft skills beyond traditional education and workplace environments. Further, while soft skills do not affect employment duration, overall skills that synergise with soft skills can be associated with longer employment durations. This suggests businesses should develop these skills via internal training or recruitment out of university for the whole set of skills relevant.

For research contributions, these results suggest businesses seeking to acquire skills should not expect to easily find workers changing jobs who have a broader array of skills. However, there are potential benefits to having deeper pools of workers groups of businesses can draw from based on the significance of certain skills at the industry level even in the short-run. For governments, these results suggest limited widespread effects of SBTC and RBTC on workers or skill categories outside of overskilling risk to a limited degree, suggesting skills development needs to be tailored for specific uses rather than anticipating broader applications to worker security. However, as these data were all before the advent of widely utilised artificial intelligence (AI), there is the need to consider changes among these associations as applications of this emerging technology develop.