Using administrative data to gain insights into microdrivers of productivity

Abstract

Since the late 1990s, the information technology revolution has improved our ability to process, store and analyse large administrative datasets. The idea of making better use of administrative data for social and economic research is not new. While researchers recognise that there is rich information contained in administrative data, they also acknowledge the need to address the methodological challenges of analysing it for research. This research explores methods to address some of these challenges and harness the opportunities from analysing administrative data to understand micro-drivers of productivity.

We start by providing the rationale for this research. We describe how we measure productivity, define firm participation in business networks and handle missing data for this research. Missing data is a common problem in data analysis, but the problem is exacerbated when we analyse an integrated administrative dataset. We need to address this problem because it can affect the conclusions that we draw from analysing the dataset. While there is no single method to address missing data, we prefer using approaches that minimise information loss.

We then address the computational challenge while studying the contributions from firm dynamics - that is, firm entry and exit, within-firm growth and reallocation - to aggregate productivity. Our experimental integrated administrative dataset contains more than 10 million workers across 1.5 million firms. We estimate unique worker- and firm-specific effects using a scalable approach. We use the estimated labour component to study firm dynamics' contributions to Australia's productivity growth in the 2002-03 to 2012-13 period across 18 industries. In general, firm exit contributes positively to productivity growth, whereas firm entry generally contributes negatively.

Firm dynamics are not the only micro-drivers of productivity. Therefore, we analyse another experimental dataset, integrating data from the Australian Bureau of Statistics, Intellectual Property Australia and Australian Stock Exchange to understand the relationship between firm participation in business networks and firm performance. We explore three types of business networks - research and development, commercial and shared directors under three different assumptions. We show the possibility of using administrative data to mitigate self-response bias in surveys. In general, we find positive associations between firm performance and these three types of business networks. Our finding is consistent with other research that draws on survey data.

We then demonstrate the possibility of using a semantic web approach to extract complex business network information from integrated administrative data. We use the business network information in the statistical network models to describe the factors contributing to firm participation in Australian business networks. We combine different sampling approaches to overcome computational problems for the statistical network models. We find that larger firms are more likely to form business networks than small and medium-sized firms. We also find that firms with more products are more likely to form business networks.

Finally, confidentiality is important to make administrative data more accessible for research. Australian businesses in some industries are characterised by an oligopoly or duopoly. Therefore, protection techniques such as information reduction may not be as effective for business microdata. We explore synthetic data to make business microdata more accessible for research while maintaining confidentiality. We find that synthetic data could be a possible approach to make business microdata more accessible for research. Show more