The idea that the world is a small place has been around for a long time. In the 1960’s the psychologist Stanley Milgram, who is perhaps best known for his controversial work on obedience behavior (Milgram, 1963), conducted an experiment to determine the average path length for social networks of people in the United States of America (Milgram, 1967). His findings indicated that on average people in the United States were separated by 5.5 friendship links. Since then, the idea of 'six degrees of separation' has achieved widespread recognition, and has given rise to plays, films, and games. The idea has also achieved a measure of perceived prominence amongst mathematicians with the 'Erdos Number' (Goffman, 1969).
On a more serious note, the small-world network phenomenon has been the subject of a number of studies, and has implications for a range of issues from the robustness and efficiency of transportation and power networks to models of neural networks (Watts and Strogatz, 1998). More recently, attention has focused on the Internet with a recent study (Leskovec and Horvitz, 2008) of the Microsoft Messenger instant messaging (IM) network finding that the average path length among Messenger users is 6.6.
There is also a significant body of work endeavoring to explain how knowledge is created and diffused through collaborative networks. Knowledge creation occurs when new information is integrated into the network or when the existing information within the network is recombined in new ways. A long line of research emphasizes the latter method, suggesting that the creation of new knowledge is most often the result of novel recombinations of known elements of knowledge, problems, or solutions (Schilling and Phelps, 2004). Much of this work has focused on patent registration data to proxy collaboration and knowledge. Investigations indicate that the existence of a tie is found to be associated with a greater probability of knowledge flow, with the probability decreasing as the path length (geodesic) increases (Sing, 2005).
Work has also been done to estimate measures of human capital by making use of linked employer-employee data from the US (Abowd, Lengermann, and McKinney, 2003). In New Zealand, there is considerable interest in anything that can help productivity in general and labour productivity in particular. Given that knowledge creation and diffusion can be said to enhance efficiency and performance, and that employee networks can be an enabler of this diffusion, understanding the characteristics of the New Zealand labour market is an important first step in developing initiatives to enhance the performance of the New Zealand economy.
Using Statistics NZ's Linked Employer-Employee Dataset (LEED), I constructed an approximation of a 'knowledge network' of wage and salary earners in New Zealand. The network spanned the period 1999–2008 in an initial attempt to understand the structure and characteristics of a network view of the New Zealand labour market.
Purpose
Recently published analysis of messages sent over the Microsoft Messenger instant-messaging network has shown that the old maxim of six degrees of separation is not far from the truth. The idea is that, on average, you are connected by no more than six links to all other 6.7 billion people on Earth. These links can be through blood, friendship or an acquaintance – you know someone who is friends with someone whose sister is married to someone ... and so on. Using Statistics New Zealand's Linked Employer-Employee Dataset (LEED), this maxim is tested on a network of wage and salary earners in New Zealand.
Methodology
This paper uses Statistics NZ’s Linked Employer-Employee Data (LEED) over the nine-year period April 1999 – March 2008 to create a 'knowledge network' of wage and salary earners in New Zealand. This network is then used to calculate the average shortest path between wage and salary earners, together with a range of measures which describe characteristics of this unique view of the New Zealand labour market.
Key Results
The network of wage and salary earners in New Zealand displays attributes of a small-world network, with any two random people able to be connected within four steps on average. The largest connected component of the network encompasses over 98 percent of wage and salary earners and appears relatively resilient to node deletion. The network displays a high degree of cliquishness.