2. Dataset
Few international econometric studies have used datasets that provide a broad view of the practices within firms. Where such studies exist (e.g., Spanos & Lioukas 2001; Bloom et al. 2005), the datasets used would not pass the exacting standards of an official statistical agency (primarily because of small sample sizes and/or low response rates). New Zealand policy agency requirements for a more sophisticated understanding of firm practices & performance led, in 2001, to the introduction of the Business Practices Survey (BPS), which surveyed a wide set of firm practices including questions on strategy, customer & supplier relations, HRM, benchmarking & quality control, together with (self-reported) performance metrics.2 The survey was designed primarily from an understanding of the management, marketing & economics literatures, with a limited number of CIS-like3 innovation questions (Knuckey & Johnston 2002).
Econometric research using this dataset produced findings that were consistent with the importance of the business practices outlined in section 1. Behaviours that were shown to be particularly important include R&D, HRM and marketing (Fabling & Grimes, forthcoming). Reinforcing the arguments above, traditional science and technology indicators were found to signal firm success, but so too were activities underlying non-technological innovation. Importantly, it was seldom the case that better performing firms only innovated on technological dimensions. The unique contribution of this research lies entirely within the survey design, which allows the contribution of specific business practices to be isolated. One finding of the Fabling & Grimes work was that many practices appear not to be significantly related to firm performance.
In 2003, SNZ ran a more CIS-consistent Innovation Survey (SNZ 2004). At present, this dataset has not been used for microeconometric research, and it was decided in 2005 that the way forward for innovation measurement in New Zealand was an integrated collection approach.4 The resulting Business Operations Survey (BOS) has a three-part modular survey design with one module focussed on firm performance (both quantitative & qualitative self-assessment) & characteristics (such as composition of employment) and two further modules examining business practices and outcomes. Though the survey runs on an annual basis, there is rotation of content yielding annual firm performance data with alternating biennial innovation and business use of ICT data (chart 3). The modular approach has been adopted for two primary purposes: first, to cope with respondent load, driven by increasing end-user needs; and, second, to enable specific policy-relevant data to be collected on an ad-hoc basis – using a third "contestable" module – without the need for a full stand-alone survey to be administered. The value of the survey is magnified by incorporating a longitudinal sub-sample so that performance can be tracked over time and relationships between practices examined across surveys. Planned linking of BOS to administrative (tax) firm performance data (IBULDD) further extends the uses of the survey.
Chart 3: Business Operations Survey modular design
→ Full size version of Chart3 [167 kB JPEG]
The BOS design process has presented an opportunity for SNZ & relevant agencies5 to bring additional perspectives on firm performance into the design process.6 As a result of changes to the innovation collection, SNZ has produced economy-wide estimates of product, process, organisational & marketing innovation under revised Oslo manual guidelines (OECD & Eurostat 2005; SNZ 2006).
BOS is a two-way stratified sample, with stratification on rolling-mean employment (RME) and two-digit industry.7 The 2005 survey was mailed to 6,979 live firms with a total of 5,595 useable responses returned (80.2% response rate). These observations are weighted to represent the population of 34,760 private, economically significant firms in NZ with six or more RMEs in all industries excluding Government Administration & Defence; Libraries, Museums & the Arts; and Personal & Other Services.
We perform two types of analysis in this paper. In section 3, we present population statistics and regressions for the BOS cross-section.8 We drop Electricity, Gas & Water Supply; and Sport & Recreation industries from the BOS dataset. This has the effect of reducing the sample size by 111 firms and the population size by 499 (1.4%), but has the advantage of putting the BOS industry coverage on a consistent basis with BPS (enabling easier comparison of the cross-section results with the panel results presented in section 4).9,10
In 2005, the contestable third module of BOS was constructed to allow direct comparison of business practice results with the BPS. In section 4 we report results for the panel of 1285 firms in both BPS and BOS. This number constitutes 46.6% of BPS responses, which is high given that BOS did not purposively survey surviving BPS respondents.11 However, there has been attrition in BPS respondents between 2001 & 2005, and there is some indication that this panel may be biased in favour of better performing firms. In particular, the primary reasons for non-availability for selection in the panel are due to firms ceasing on the statistical frame or employment dropping below the population threshold.12 This bias manifests itself through greater incidence of some "high-performance" practices in the panel relative to the BPS population. Given the importance of this issue, it is revisited in section 4. In this paper, no effort has been put into attempting to compensate for this bias and this should be borne in mind when interpreting section 4 results.13
Back to Top