• Home
• About MED
• Publications
• Occasional Papers
• Archive
• 2007
• 07/04
• 4. Panel Results

• Publications

• Occasional Papers

• Archive

• 2007

In this section we seek to relate general business practices in 2001 to innovation outcomes in 2005. To do this we construct a measure of general management practices from the subset of questions where a concordance can be confidently mapped across the BPS-BOS surveys (table 3 lists the subject areas covered).²² Specifically we regress self-reported "high" relative productivity (a binary) on population-weighted BPS practices and use the predicted probabilities generated by this model as our measure of management practices in both 2001 & 2005 (i.e., for the 2005 management practice index we use 2001 model coefficients with 2005 variable values).²³

Note: For questions with multiple response categories, each category response is entered as a separate binary variable. Where applicable the "don't know" category of each question is dropped (to avoid perfect multicollinearity), failing that the "no" category is dropped.

Before discussing the properties of this index, we revisit the issue of panel bias. In table 4 we regress a binary of whether the firm is in the panel on the 2001 (population-weighted) characteristics of firms in the BPS dataset.²⁴ Whether the firm has survived on the business frame is, naturally, a critical determinant of the panel composition. Since the survival variable captures the attrition effect, the importance of firm size should be interpreted as being related to sampling, specifically by: reducing the probability of dropping below the BOS firm size threshold; and increasing the probability the firm will be in a stratum with a higher sampling proportion. To some extent bias is also suggested by the 2001 practices & outcomes of firms in the panel. Testing across product & process innovation outcomes, innovation practices, and our management practice index, the panel has a significantly larger proportion of firms that were marketing the introduction of new products (p=0.020) & well-managed firms (p=0.000), as measured by our practices index. This latter effect is demonstrated in chart 7, where the distribution of the 2001 management practices of the panel is shown relative to the BPS population.

Note: Probit regression with selection in the panel as the dependent variable. Stars denote significance at 5% (*) & 1% (**) level (two-sided test – robust p-values in square brackets below coefficients).

Chart 7: Management practices of panel relative to BPS population in 2001

→ Full size version of Chart 7 [99 kB JPEG]

Bearing this potential bias in mind, we perform two common-sense tests on our management practices index. First we look at the persistence of practices over time, noting that there has been a general increase in the index from 2001 to 2005 (chart 8).²⁵ While practice changes are quite diverse, prior practices do play an important role in explaining current practices (panel (1) of table 5). Panel (2) of table 5 confirms that firms that reported organisational/managerial innovation over 2003-2005 also experienced changes in practices as measured by the change in our index over 2001-2005. Taken together, the regressions in table 5 suggest that our conservative approach to variable concordance is satisfactory, and that our management practice index is measuring something consistent with the respondent's sense of how the firm has changed. Panel (2) could also be considered a useful "reality check" on the respondent's interpretation of the organisational/managerial innovation question.

Note: Panel (1) is a linear regression with management practice (2005) index as dependent variable. Panel (2) is a probit with reported organisational/managerial innovation as the dependent variable. Both regressions contain ANZSIC division dummies (coefficients not shown). Stars denote significance at 5% (*) & 1% (**) level (two-sided test – robust p-values in square brackets below coefficients).

Chart 8: Change in management practices 2001-2005

We conclude this section by asking, what impact do 2001 business practices have on 2005 innovation outcomes? We test this in two stages (table 6). First we test the impact of innovation practices & sources of information on innovation groups. We then introduce our measure of how well managed the firm is in 2001 to see what impact this additional explanatory variable may have. Focussing on panel (1), very few innovation activities are found to be significantly linked to positive innovation outcomes four years on. Consistent with our concerns about the potential for returns to R&D being lagged, both conducting in-house R&D & engagement with universities/polytechnics are now positively associated with innovation, whereas contemporaneously they were not. Marketing new products stands out as being positively associated with innovation outcomes both contemporaneously & across time. Other innovation activities – machinery & equipment investment, design, innovation-related employee training, show no longer term relationship with innovation but all had strong positive contemporaneous relationships (table 2), perhaps suggesting that the intent of these activities is more to enable the production of current innovations rather than as investments in the future innovative capacity of the firm.²⁶

Note: Both panels are multinomial probits with innovation group as the dependent variable (NON is the base outcome). Regressions contain ANZSIC division dummies (coefficients not shown). Stars denote significance at 5% (*) & 1% (**) level (two-sided test – robust p-values in square brackets below coefficients). There are no significant (5% level) differences between PP & OM innovator coefficients in either panel. Proportions of each innovation group accurately predicted are shown below the table.

Finally, in panel (2), we introduce 2001 management practices. Our punch-line is twofold: the initial endowment of management practices is good for (COMBO) innovation outcomes in 2005; and the introduction of the management variable does not have substantial impact on the coefficients or significance of the importance of in-house R&D & marketing of new products. In other words, we find that good management practices appear to provide an additional effect on future innovation outcomes.