1504.0 - Methodological News, Mar 2011  
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 31/03/2011   
   Page tools: Print Print Page Print all pages in this productPrint All

Modelling the Relationships between Government Assistance, Innovation, R&D and Business Performance

The Analytical Services Unit (ASU) is currently involved in a collaborative work between ABS and the Department of Innovation, Industry, Science and Research (DIISR), on modelling the relationships between government assistance, innovation, R&D and business performance.

The study utilises four years of data from the ABS Business Longitudinal Database (BLD), which has detailed information on government assistance, business demographics, innovation activity, ICT usage, financial information, and many other variables relevant to the analysis.

The investigation aims to provide information on the modelled relationships between government assistance (in the form of grants, funding, subsidies, tax concessions or rebates) and innovation; the relationship between innovation and business performance (e.g. turnover, value added); the likelihood of innovation in the absence of government assistance; the changes in the relationship between government assistance and innovation over time; and the different factors that may be associated with the probability of a firm innovating, after receiving government support. Some of these factors include R&D, age of the firm, the industry where it operates, location, business size and ICT intensity.

ASU's approach to the analysis involves the testing of the usefulness of Propensity Score Matching (PSM), which is an applied technique used in policy evaluation settings. PSM matches participating firms, such as the firms that received government assistance, to firms which did not, but which have similar characteristics. It uses conditional probabilities to compute a score for each of the participating and non-participating firms, usually by using logistic or probit modelling. Once the scores are computed, a matching algorithm is used --such as nearest neighbour, radius/caliper, kernel or local linear-- to match the participating firms with the non-participants. The matched firms are then used in the analysis of the impact of a treatment (i.e. government assistance) on the outcome (e.g. innovation). The PSM is attractive in that it addresses the selection bias which is often a problem in non-experimental settings. Selection bias in this study refers to the difference between the counterfactual for participating firms and the observed outcome for non-participating firms. By implementing a PSM, ASU expects to get more statistically robust and accurate results.

For further information, please contact Franklin Soriano on (02) 6252 5933 or franklin.soriano@abs.gov.au