Bookworms and party animals: an artificial labour market with human and social capital accumulation

Abstract

Data show that educated workers earn higher wages and are unemployed less often. Some researchers believe that education improves a worker's productivity (or human capital), making them more desirable on the job market, while others believe that it improves a worker's network (or social capital), giving them more information about lucrative openings and more resources to secure a job (such as references from peers). Much of the research on human and social capital focuses on quantifying the various impacts of schooling on workers and often overlooks how economic systems actually manage to produce those outcomes. This paper develops an agent-based complex adaptive system featuring formal schooling and on-the-job training, social networks, labour market search and durable employment contracts to explain the process linking education to labour market outcomes and economic performance in New Zealand (and similar economies). Sample simulations show that human capital accumulation explains many of the novel facts seen in the data, while social capital alone is not enough.

Keywords:

• agent-based modelling
• human capital
• social capital
• economic aspects of education.

Acknowledgement

Special thanks to Viktoria Kahui, David Fielding, the attendees of the University of Otago Economics Department Seminar Series, participants at the 13th Annual Conference of the International Network for Economic Research (INFER), and two blind reviewers for contributing to the improvement of this paper. All errors are my own.

Notes

1. For example, human capital is often measured as ‘years in school’ which does not necessarily account for the value of the particular information learned in different fields or programs (Schuller and Field 1998).

2. Many of these models are steeped in the rational choice paradigm. For example, we often ask an agent to consider his entire discounted lifetime income when choosing how much schooling to buy – a bit much to ask for from an agent who is currently uneducated.

3. This is known as a generative approach to social science. According to Epstein (1999), generative approaches are more able to truly explain phenomena than inductive and deductive methods.

4. With simulation, we can analyse the potential impact of policies for which we have no data (i.e. policies which have never been implemented).

5. More information about NetLogo in addition to the software itself is available at http://ccl.northwestern.edu/netlogo (Wilensky 1999).

6. Data on the New Zealand economy include a particularly deep recession from 1990 to 1992. Limiting the sample to 1993–2007 brings the average growth rate of GDP per capita to 2.4% [st.dev. 1.3%].

7. Note that the simulation program calculates the current tuition level as a fraction of the average uneducated wage that prevailed in the previous period. Ideally, financial data from New Zealand universities for 2010 would be used to compute this fraction, however these data were unavailable at the time this project was conducted and the above approximation was used. A recent data release from the Ministry of Education – New Zealand (2011) provide revenue data for 2010 and suggest a calibration of 89%. The difference between this value and the approximation above is small and has little impact on the model implications.

8. Because colleges accept a fixed fraction of applicants in the model, this acceptance policy implies that fluctuations in enrolments are driven principally by fluctuations in applicants and not the college’s willingness to admit students.

9. The wage gap worsens as the economy develops and the effects of the initial population are ‘filtered out’. This process however takes 70 periods. For the ending periods of the program, the wage gap fluctuates around a stable level of 63.4%.

10. While educated workers are frequently employed more often than uneducated workers, they are not always employed more often (as shown in the last few periods in Figure 9). The nature of the simulation is rich enough to allow the unemployment rate of educated workers to exceed that of uneducated workers if the economy demands it. For example, if a large number of students were to graduate simultaneously, the unemployment rate for educated workers may be pushed up depending on how their skills measured up against more experienced uneducated workers.

11. Note that the posted tuition (set as a fraction of average unskilled earnings) and the expected value of a potential student’s outside option (becoming an uneducated worker) rises, which provide less incentive for students to apply for college (a ‘substitution effect’). Simulations suggest that the ‘income effect’ associated with college becoming more affordable seems to dominate and the number of incoming students rises above the number of graduating students.

12. As before, workers engaged in on-the-job training learn at the rate θ_b = 0.20.

13. This experiment tests whether or not the model produces outcomes similar to those described by Buerkle and Guseva (2002).

14. The unemployment rate for bookworms is on average 0.3% [st.dev. 0.9%] less than party animals in this simulation which is not statistically different from 0.

15. The unemployment rate for bookworms is on average 0.6% [st.dev. 1.7%] greater than party animals in this simulation which is not statistically different from 0.

16. These results refer to all types of tertiary institutions. The International Standard Classification of Education (ISCED) distinguishes between Type B institutions (practical, technical or skill-based programmes with a minimum duration of 2 years full-time equivalent) and Type A/Advanced Research Programmes (theory-based programmes with a minimum duration of 3 years full-time equivalent). For New Zealand, the implications made by Table 4 differ substantially from those made in Table 1. Table 1 is generated using more detailed earnings data and incorporates earnings from all sources.

17. Note, however, that this does not preclude the need to produce unique agent-based frameworks for Australian and British economies separately.