Thriving through innovation: lessons from the top Acknowledgement CSIRO acknowledges the Traditional Owners of the lands that we live and work on across Australia and pay their respect to Elders past, present and emerging. CSIRO recognises that Aboriginal and Torres Strait Islander peoples have made and will continue to make extraordinary contributions to all aspects of Australian life including culture, economy and science. The project team would like to thank Damian Hine for early guidance and Tim Kastelle, Mark Dodgson, Martie-Louise Verreynne, Michael Rosemann, Mark Bazzacco and Jamie Ford for their particularly helpful feedback on early drafts of this report. Contact us 1300 363 400 csiro.au/contact csiro.au For further information Dr Jerad Ford +61 7 3327 4216 jerad.ford@csiro.au csiro.au/futures Citation Ford JA and Brea E (2020). Thriving through innovation: lessons from the top. CSIRO, Australia. Copyright © Commonwealth Scientific and Industrial Research Organisation 2020. To the extent permitted by law, all rights are reserved, and no part of this publication covered by copyright may be reproduced or copied in any form or by any means except with the written permission of CSIRO. Important disclaimer CSIRO advises that the information contained in this publication comprises general statements based on consultations and desktop research. The reader is advised and needs to be aware that such information may be incomplete or unable to be used in any specific situation. No reliance or actions must therefore be made on that information without seeking prior expert professional, scientific and technical advice. To the extent permitted by law, CSIRO, the Steering Committee and Sponsors and Supporters (including its employees and consultants) excludes all liability to any person for any consequences, including but not limited to all losses, damages, costs, expenses and any other compensation, arising directly or indirectly from using this publication (in part or in whole) and any information or material contained in it. CSIRO is committed to providing web accessible content wherever possible. If you are having difficulties with accessing this document, please contact csiro.au/contact. 2 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland Research design In order to identify innovative behaviour associated with superior performance in Australian businesses, we searched for patterns of innovation activity across the three dimensions inspiring our investigation: (1) what are companies doing to innovate; (2) what processes are companies following to innovate; (3) what is the companies’ attitude towards innovation. We used multiple measures for each dimension, adopted from the most influential literature on innovation management from academia ❶ and practice. Approach While we relied on survey-based research to collect data on innovation activities from executives from Australian businesses, we also relied on secondary data in order to make robust observations about the performance effects of these innovation activities. The use of secondary data that measure performance in absolute terms represents an advancement over the typical survey research that collects relative performance data from the survey itself. The analysis of the data and results are supported by cutting-edge methods including multiple statistical tests and predictive modelling. Publicly listed companies offer a proper setting to systematically investigate innovation management and strategy behaviour given the amount and type of company data available resulting from the obligations tied to public ownership. More data equals more powerful and robust insights. But also, public companies in Australia represent a diverse set of companies comprising successful world-class companies, influential players in the domestic scene, and mature companies that have been key to the economic stability of the nation. Taken together, the ASX is a top-performing group compared to other listed exchange groups in the world. This makes it an appropriate place to look for pockets of successful innovation activity and learn from them. ❷ We took a snapshot of the ASX in November 2019 (which included 2,195 companies) and searched from financial data in large proprietary Sampling company financial datasets. After refining our sample to include those with available accounting data for at least three years, we obtained a final sample of 197 companies with complete survey results. We analysed the representativeness of our sample compared to the entire population of ASX-listed companies, and the distributions across GICS (Global Industry Classification Standard) sectors and company size (by market capitalisation) for both sample and population are similar. 3 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland National survey ❶ Design The national survey was designed so that it captured the most relevant features of the innovation activities carried out by the companies in our sample. The targeted respondents were c-suite executives, managing directors, vice-presidents, directors and managers from corporate units responsible for undertaking company-level innovation efforts. We conducted a systematic review of the innovation literature and identified constructs that have been employed to capture innovation activities across our “pattern, process and attitudes” dimensions, and with evidence of positive associations with business performance. We then identified a series of survey scales that have been designed, tested and successfully employed by leading academics and practitioners to measure our constructs of interest. ❷ Development The survey questions were carefully developed so that the responses could clearly reflect the constructs of interest. Some questions were presented in the form of Likert scale-type questions, others were presented as yes/no questions, while others were presented as closed-ended direct questions. We standardised the coding structure across questions to maintain consistency, particularly in the Likert scale questions. The questions, scales and underlying constructs of interests were validated by a panel of experts from The University of Queensland. We also discussed the survey design with key managers from the Department of Industry, Science, Energy and Resources. We then conducted internal survey assessments, where we checked for consistency, interpretability and duration. Lastly, we teamed up with a multinational market research firm, who conducted additional checks to the survey program via pilot-testing. ❸ Administration For the administration of the national survey, we partnered with a world-leading multinational market research and consulting firm. In addition, we worked with multiple vendors to source and validate the contact details of the potential respondents. The survey asked 10 multi-item questions, with an average duration of 16 minutes. It was administered via phone calls to the targeted respondents. A snow-balling technique was used in cases where the specific respondent was unable to participate but the organisation was. The survey remained active for 2 months, from late January 2020 to late March 2020. During this period, COVID-19 emerged as a global pandemic and disrupted the economic activities in Australia. Undoubtedly, an unprecedented situation like this had an effect on the responsiveness and willingness to participate. Despite this, we managed to obtain a substantial number of complete surveys on time. 4 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland Secondary data, integration and processing Secondary data collection We collected financial data and corporate details for our sample companies from Morningstar’s DatAnalysis and S&P’s Capital IQ, two of the most comprehensive financial data sources for listed companies in Australia. We gathered corporate details such as location, industry of operation and listing date, as well as longitudinal data comprising a variety of financial indicators reflecting the financial performance of the companies. Data storage, integration and processing After pre-processing, primary (survey) data and secondary data were stored in robust relational databases purposefully designed for the study. For each company, primary and secondary data were integrated using the company ticker. The data cleaning, processing, and analysis was conducted in R, one of the most comprehensive software environments for statistical computing and graphics. 5 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland Innovation variables (independent variables) Each of the 10 main questions in the survey (comprising a total of 85 items) captured a particular aspect of the innovation activity in ASX companies. These aspects were directly associated to each of our three dimensions of analysis (innovation patterns, innovation processes and attitudes toward innovation). A total of 13 innovation variables (and one self-assessed innovation performance variable) were constructed by aggregating multiple items. We employed different aggregation methods -ranging from sum to unweighted and weighted averages, and all variables were normalised. These procedures mirrored those indicated in the publications from which we sourced the scales. We checked the reliability and internal consistency among the items by conducting principal components analysis (PCA) and calculating Cronbach’s alpha. The majority of the individual factors across the 13 variables had factor loadings greater than 0.5 (those below this number were discarded). All the 13 variables had a Cronbach’s alpha of 0.7 or higher, values that are similar to those reported in the reference sources. See next slide for the full list of question, items and derived variables, with corresponding references. 5 variables Innovation survey es4 6 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland Innovation variables (cont.) Questions (with references) Items Scale Variable 1. Is your business actively usingthe following technologies to create value for the business and/or for your customers?1 2 3 2. In the last 3 years, has your business introduced any of the following types of innovation? For each one, what was the degree of novelty?4 3. Rate how well these activities have supported the introduction of new products and services in the past 3 year.5 Robotic Hardware | Robotic Process Automation | Autonomous vehicles | Augmented / Virtual Reality | Internet of Things | Advanced data analytics | Artificial Intelligence | Cognitive computing | Conversational computing | Mobile computing | Cloud / edge computing | Blockchain | Digital twins | 3D printing | Advanced materials | Renewable energy | Energy storage | Carbon abatement | Space technologies | Advanced biotechnologies Product innovation | business process innovation | business model innovation New to the world | New to Australia | New to the industry within Australia Acquisition of R&D | Acquisition of machinery, equipment or software | Acquisition of externalknowledge through intellectual property | Acquisition of knowledge from other organisations | Acquisition of knowledge through innovation contests | Participation in external innovation programs In-house R&D | Training of staff | Marketing | Design work Dichotomous (yes, no) Dichotomous (yes, no) Dichotomous (yes, no) 5-point Likert scale 5-point Likert scale Breadth of technology use Innovation types Degree of novelty Degree of technology and innovation acquisition Degree of in house innovation activities * Items dropped after internal consistency checks Patterns Processes Attitudes Alternative performance variable 7 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland Innovation variables (cont.) Questions (with references) 4. Please answer these questions about your innovation performance and management.6 7 5. How important have these sources of information been for your innovation activities in the last 3 years?6 6. Please rate the effectiveness of these mechanisms to protect and derive value from new products and services developed by your business.6 8 9 7. Please indicate how often each of the following items is employed when undertaking innovation activities.11 Items Percentage of revenues from products and/or services that are new to the business in the past 3 years Number of innovation projects that have been funded in the past 3 years Number of innovation projects that have been stopped or discontinued in the past 3 years Other enterprises within your enterprise group* | Suppliers of equipment, materials, components, or software | Clients or customers from the private sector | Clients or customers from the public sector | Competitors or other companies in your sector | Consultants or commercial labs | Universities or other higher education institutes | Government, public or private research institutes Patents | Design registration, such as trademarks, copyrights and servicemarks | Secrecy | Lead time | Complexity of product or design| Complementary service provision | Complementary marketing capabilities | Complementary production capabilities Employees are regularly rotated between jobs in our business | There is regular talk about possibilities for collaboration between units | Our business coordinates information sharing between units through a knowledge network | We have cross-functional teams to exchange knowledge between departments | Our business uses temporary workgroups for collaboration between units on a regular basis In our business, there is ample opportunity for informal “hall talk” among employees | In our business, employees from different departments feel comfortable calling each other when the need arises | People around here are quite accessible to each other | In our business, it is easy to talk with virtually anyone you need to, regardless of rank or position Scale Variable Continuous Integer Integer 5-point Likert scale 5-point Likert scale 5-point Likert scale 5-point Likert scale Innovation performance Portfolio management strategy Breadth of external knowledge search Strength of IP protection mechanisms Use of cross functional interfaces Use of network structures * Items dropped after internal consistency checks Patterns Processes Attitudes Alternative performance variable 8 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland Innovation variables (cont.) Questions (with references) Items Scale Variable 8. Please rate how well the following statements describe the innovation initiatives in your business.12 9. Has your business cooperated on innovation activities with any of the following types of organisation in the last 3 years? 6 10. Please rate your business's strategic posture by rating the following statements 13 Employees cooperate fully in generating and screening new ideas for new products and services (reversed)* | Top management has the exclusive task of establishing goals and priorities for our innovation strategies | The innovation process is heavily formalized through dedicated projectteams and routinized procedures | In our business, we have a dedicated unit responsible for the innovation function Other enterprises within your enterprise group* | Suppliers of equipment, materials, components, or software | Clients or customers from the private sector | Clients or customers from the public sector | Competitors or other companies in your sector | Consultants or commercial labs | Universities or other higher education institutes | Government, public or private research institutes In general, the top managers of my business favour a strong emphasis on R&D, technological leadership and innovation | We have very many new lines of products or services that have been marketed in the last 3 years | Changes in our product or services lines have usually been quite dramatic over the last 3 years | My business typically initiates actions which competitors then respond to | In dealing with competitors, my business is very often the first business to introduce new products/services, administrative techniques, operating technologies, etc. | My business typically adopts a very competitive “undo-the-competitors” posture | Our business has a strong proclivity for high-risk projects with chances of very high returns | In general, the top managers of our business engage in bold, wide ranging acts that are necessary to achieve our objectives | Inuncertain times, my business typically adopts a bold aggressive posture in order to maximise the probability of exploiting potential opportunities. 5-point Likert scale Dichotomous (yes, no) 5-point Likert scale Degree of process formalisation Breadth of collaboration Degree of entrepreneuria l strategic posture * Items dropped after internal consistency checks Patterns Processes Attitudes Alternative performance variable 9 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland Performance variables Performance variables (dependent variables and control variables) For the financial performance measure (dependent variable), we chose a market-based measure of return instead of profitability ratios such as return on assets, given its forward-looking nature and capacity to capture the company’s future performance potential.14 This is particularly important in innovation studies, as the bottom-line effect of novel products and services take time to materialise. The way market-based measures such as market-to-book ratio and Tobin’s Q reflect the value of the company’s future earning and growth potential is by combining the tangible value of a company with the intangible value that is perceived by the market. Evidence suggests that the market is capable of evaluating innovative activity in firms reasonably well.15 We adopted Morningstar’s formula for calculating price-to-book value, computed as the closing share price on the last day of the financial year divided by shareholders equity per share. We then calculated the median of the price-to-book value for each company for the last three years. Our choice of performance variable and its calculation help us ensure that there is no significant correlation between the dependent variable and control variables used in the study, particularly firm size (measured by total assets). In our models, we controlled for the differences in innovation activity across industries by introducing dummy variables for the two- digit GICS codes in which the company operates.8 Given the potential effects of company size on innovation (in terms of resources available for innovation and complementarities between marketing and finance capabilities and innovation), we also employed a control variable for company size, calculated as the median of the natural logarithm of total assets for the last three years.16 For all variables, the three-year window helped us maintain consistency with the survey questions, which asked about the innovation activities performed in the last three years. Due to the presence of extreme outliers, the dependent variable was winsorised at 5% and 95%. Winsorisation is a technique used to treat highly skewed variables that may lead to biased results when conducting statistical analyses, and consists of replacing extreme observations by percentiles within the specified cut-off range.17 10 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland Final variables list Variable Description and measurement Source Survey variables Processes Degree of engagement in activities where tangible and intangible elements associated with innovation were developed Degree of in-house in-house. Measured by the mean scores of activities conducted by and within the business that have supported the 5 innovation activities introduction of new products and services in the past 3 years (normalised). Whether the company follows a project portfolio management process in which innovation effort and resources are Portfolio management selectively allocated across a number of projects. This dichotomous variable takes to value of 1 (“selective”) if the 7 strategy number of both projects funded and stopped in the past 3 years are greater than 0, and 0 (“unselective”) otherwise. Breadth of external Degree of use of external sources of information and knowledge by the company in its innovative activities in the last 3 6 knowledge search years. Measured by the sum of scores for each knowledge source (normalised). Extent to which the innovation processes are generated by managers, and formally structured through routines, Degree of process frames and dedicated units across the company. Measured by the weighted mean of scores of the corresponding 12 formalisation items, using the item factor loadings as weights (normalised). Degree to which the company has cooperated with different types of organisations as part of their innovation activities Breadth of collaboration 6 in the last 3 years. Measured by the sum of scores for each type of collaborator organisation (normalised). 11 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland Final variables list (cont.) Variable Description and measurement Source Survey variables Patterns Breadth of technology use Number of different types of technologies employed to create value for the company and its customers. Measured as 18 the sum of the numbers of technologies used (normalised). Number of innovation types (i.e. product, process, business model) simultaneously introduced in the last 3 years. The Innovation types 19 possible values of this categorical variable are 1 type, 2 types, or 3 types. Average newness of the innovations introduced in the last 3 years, based on four levels of novelty: (1) new to the Degree of novelty business, (2) new to the industry (within Australia), (3) new to Australia and (4) new to the world. Measured as the 20-21 mean of the levels of novelty across the different types of innovation introduced (normalised). Degree of engagement in activities where tangible and intangible elements associated with innovation were purchased Degree of technology and or acquired from external parties. Measured by the mean scores of acquisition-based activities that have supported 5 innovation acquisition the introduction of new products and services in the past 3 years (normalised). 12 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland Final variables list (cont.) Attitudes Variable Strength of IP protection mechanisms Use of cross-functional interfaces Use of network structures Degree of entrepreneurial strategic posture Description and measurement Source Survey variables Degree of use of informal and formal IP protection methods to derive value from the new products/services 6,8-10 developed. Measured by the sum of scores for each IP protection method (normalised). Degree of use of cross-functional integration mechanisms (such as cross-functional teams or projects) when innovating. It also reflects the degree of reliance on formal organisational structures to innovate. Measured by the 11 weighted mean of scores of the corresponding items, using the item factor loadings as weights (normalised). Degree of use of employee networks within the company as integration mechanism when innovating. It also reflects the degree of reliance on informal organisational structures to innovate. Measured by the weighted mean of scores of 11 the corresponding items, using the item factor loadings as weights (normalised). Extent to which the company’s strategic posture is characterized by frequent technological and product innovation, an aggressive competitive orientation, and a strong risk-taking propensity by top management. Measured by the mean 13 scores of the items of the three traits used to assess strategic posture: innovation tendency, proactiveness and risk- taking propensity (normalised). 13 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland Final variables list (cont.) Variable Description and measurement Source Secondary data variables Market-based indicator of the long-term financial performance of the company. It integrates both the intangible and tangible value of the company, allowing to estimate future performance potential based on the present innovation Price-to-book value 15,24 activities. The measure is provide by Morningstar, and is calculated as the 3-year median of the closing share price on the last day of financial year divided by shareholders equity per share, over the last 3 years (winsorised at 5% and 95%) Indicator of the size of the company using total assets. Measured by the 3-year median of the natural logs of the yearly Total assets 16,23 total assets, over the last 3 years. Used in the models to control for the effects of size differences on performance Indicator of main industry of operation. Measured by taking the first two digits of the company GICS code. Used in the GICS sector 8,23 models to control for the effects of the industry conditions on performance. Innovation output variable that reflects the company’s capacity to translate innovation activities into products and Innovation performance services. Measured by the percentage of revenues from products and/or services new to the company in the last 3 6 years. Used for robustness checks to test the reliability of the estimations Alternative market-based indicator of financial performance. Measured by the 3-year median of the market Market-to-book ratio capitalisation divided by total assets, over the last 3 years. Used for robustness checks the assess the effect of the 14 choice of market-based performance measure as dependent variable Alternative profitability-based indicator of financial performance. Measured by the 3-year median of the fiscal year's Return on assets earnings before interest divided by total assets, over the last 3 years. Used for robustness checks to assess the effect of 22 the choice of performance measure as dependent variable Financials and other 14 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland Sample of respondents and representativeness The 197 companies that completed the survey were categorised by size (market capitalisation) and GICS industry group. The distribution across groups were compared with the distribution for the full list of more than 2000 ASX companies (taking the existing population on 5th March 2020 as baseline). Although the distributions are not identical, none of the categories are underrepresented or overrepresented by more than 8 percent point (excluding “NA” and “Not Applic”). The exception is “Materials” (over representation by 14.7 percent points). However, the industry dominance is of similar proportion in both sample groups. All in all, the distributions are moderately similar, thus, our sample seems to be a fair representation of the entire population. Distribution across GICS industry group Distribution across size (market cap.) 45 5 10 15 20 25 30 35 40 Percentage 0 5 10 15 20 25 30 35 40 Percentage Our sample All ASX 0 Small(Less than Small-mid(Bet. Mid(Bet. $17.04M Mid-large(Bet. Large (greater than $4.48M) $4.48M& $17.03M) & $55.05M) $55.06M& $267.82M) $267.82M) 15 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland -0.01 1.00 0.27 0.32 1.00 0.18 0.05 0.32 1.00 0.13 0.14 0.24 -0.04 1.00 0.14 0.09 0.42 0.48 -0.02 1.00 0.30 0.30 0.58 0.34 0.27 0.49 1.00 0.18 0.08 0.30 0.23 0.17 0.33 0.29 1.00 0.05 0.15 0.13 0.02 -0.05 0.13 0.05 0.37 1.00 0.11 0.22 0.31 0.09 0.21 0.18 0.52 0.15 -0.08 1.00 0.26 0.26 0.27 0.32 -0.01 0.48 0.30 0.26 0.12 0.16 1.00 0.35 0.46 0.15 0.35 0.28 0.47 0.32 0.14 0.33 0.32 1.00 -0.04 0.05 0.01 -0.04 0.22 0.03 0.07 0.08 0.00 0.09 0.05 0.16 1.00 0.43 -0.21 0.02 0.06 -0.03 0.05 -0.03 0.01 -0.18 0.06 0.10 -0.13 -0.16 1.00 -0.01 1.00 0.27 0.32 1.00 0.18 0.05 0.32 1.00 0.13 0.14 0.24 -0.04 1.00 0.14 0.09 0.42 0.48 -0.02 1.00 0.30 0.30 0.58 0.34 0.27 0.49 1.00 0.18 0.08 0.30 0.23 0.17 0.33 0.29 1.00 0.05 0.15 0.13 0.02 -0.05 0.13 0.05 0.37 1.00 0.11 0.22 0.31 0.09 0.21 0.18 0.52 0.15 -0.08 1.00 0.26 0.26 0.27 0.32 -0.01 0.48 0.30 0.26 0.12 0.16 1.00 0.35 0.46 0.15 0.35 0.28 0.47 0.32 0.14 0.33 0.32 1.00 -0.04 0.05 0.01 -0.04 0.22 0.03 0.07 0.08 0.00 0.09 0.05 0.16 1.00 0.43 -0.21 0.02 0.06 -0.03 0.05 -0.03 0.01 -0.18 0.06 0.10 -0.13 -0.16 1.00 Descriptive statistics and correlations There are no systematic issues with missing values. Although “Degree of novelty” has missing values for 26% of the companies, these are cases were novelty could not be computed given the lack of innovations introduced in the last 3 years, and does not represent missing information. Only two Spearman’s rank-order correlation coefficients are greater than +/-0.5. For the first one, the correlation between ”Strength of IP protection mechanisms” and “Degree of in-house innov. activities” could be due that companies with high levels of R&D, marketing and design work are more likely to require stronger IP protection strategies. For the second one, an explanation of the correlation of “Degree of process formalisation” and “Strength of IP protection mechanisms” could be that a more formalised innovation process might result on a wider selection of IP protection methods, as there is more certainty about the type of innovation outputs and their needs in terms of appropriability mechanisms. variable n mean sd median min max skew kurtosis se variable 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Breadth of technology use 197 0.26 0.19 0.25 0.00 0.85 0.80 0.27 0.01 1 Breadth of technology use 1.00 Innovation types* 197 2 Degree of novelty Degree of novelty 145 0.71 0.28 0.75 0.25 1.00 -0.41 -1.26 0.02 3 Degree of in-house innov. act. Degree of in-house innov. act. 197 0.45 0.16 0.47 0.14 0.72 -0.23 -0.89 0.01 4 Degree of tech. and innov. acq. Degree of tech. and innov. acq. 197 0.31 0.12 0.31 0.12 0.64 0.27 -0.53 0.01 5 Innovation performance Innovation performance 184 0.20 0.32 0.00 0.00 1.00 1.53 0.88 0.02 6 Breadth of ext. knowledge search Portfolio management strategy* 191 7 Strength of IP protection mech. Breadth of ext. knowledge search 196 0.52 0.16 0.50 0.14 0.91 0.31 -0.55 0.01 8 Use of cross-functional interfaces Strength of IP protection mech. 196 0.53 0.21 0.55 0.08 0.98 0.06 -0.89 0.02 Use of cross-functional interfaces 197 0.43 0.17 0.45 0.15 0.76 -0.21 -1.06 0.01 9 Use of network structures Use of network structures 197 0.70 0.15 0.75 0.17 0.84 -1.60 2.80 0.01 10 Degree of process formalisation Degree of process formalisation 197 0.41 0.17 0.41 0.16 0.78 0.35 -0.70 0.01 11 Breadth of collaboration Breadth of collaboration 197 0.54 0.29 0.57 0.00 1.00 -0.20 -0.96 0.02 12 Degree of entrep. strategic posture 0.23 Degree of entrep. strategic posture 197 0.60 0.18 0.60 0.20 0.98 -0.17 -0.64 0.01 13 Price-to-book value 197 2.92 3.32 1.83 -0.67 12.38 1.68 2.14 0.24 Price-to-book value 14 Total assets (log) Total assets (log) 197 17.20 2.63 16.56 10.83 25.98 1.07 1.07 0.19 GICS sector* 197 * Categorical variables 16 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland Quantile regression analysis After confirming that there were no significant correlations between the constructed variables, we estimated quantile regression models for each the ‘pattern, process and attitudes’ dimensions in our study. The dependent performance variable, price-to-book value, was regressed on each set of the innovation variables comprising the three dimensions (and controlling for size and industry), resulting in three separated models. We conducted separated analyses because we wanted to assess the performance effect of innovation patterns, processes, and attitudes irrespective of each other. Although we do not rule out the hypothesis that these three different innovation dimensions interact with each other to drive performance, assessing such interaction is out of the scope of this study and may unnecessary add complexity to our quantile regression models. We chose quantile regressions over more traditional linear regressions for two reasons. Firstly, the dependent variable follows a non-normal and highly skewed distribution, which is problematic for traditional linear regression. Quantile regression is well equipped to deal with uneven distributions of the independent variables across the different quantiles of the dependent variable.27 Secondly, we seek to understand the different ways in which innovation affects performance in companies across different performance levels – doing so maximises the likelihood of finding innovation behaviour that is far from average, generating insights that could encourage changes in the innovation behaviour of average companies in Australia. This is precisely the reason why quantile regression has been employed in influential innovation studies in the past.15 25 26 In our quantile regression analyses we make comparisons between the 90th and the 50th percentile. This comparison is meant to reflect the difference between the average ASX performer—what we term as mid-performers—and those at the top—i.e. the 90th percentile. This comparison is also a more conservative approach than comparing the top with the bottom performers. However, it represents a fairer comparison given that these companies may be in the lowest performance category for many other reasons outside of the study’s scope, which could confound our analysis. 17 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland Quantile regression results ❶ Patterns Innovation patterns: Quantile regression results for 10th, 25th, 50th, 75th and 90th quantiles of the dependent variable (Standard errors of coefficients in parentheses. Obtained using 200 bootstrap replications) Price-to-book value OLS Quantile regression 10th 25th 50th 75th 90th Degree of novelty 1.15 -0.21 -0.19 -0.84 1.07 7.76*** (1.10) (0.55) (0.64) (0.76) (1.97) (2.93) Number of innov. types: 2 types 0.01 -0.34 -0.39 -0.59 0.2 0.99 (0.74) (0.46) (0.55) (0.62) (1.84) (2.02) Number of innov. types: 3 types 1.33 -0.4 -0.09 0.07 1.92 4.12* (0.92) (0.70) (0.65) (1.01) (1.92) (2.30) Breadth of technology use -0.84 -1.26 -0.21 1.62 -2.66 -5.84 (1.93) (1.21) (1.31) (1.76) (3.36) (4.44) Degree of tech. & innov. acquisition -1.13 0.35 -1.18 -0.07 0.28 -1.11 (2.68) (1.53) (1.78) (2.74) (5.48) (7.28) Control variables: Company size (log total assets) -0.04 0.18* 0.07 -0.1 0.01 -0.17 (0.13) (0.10) (0.11) (0.14) (0.29) (0.34) GICS sector (dummies) Included Included Included Included Included Included Constant 2.68 -2.63 0.05 3.38 5.26 3.31 (3.42) (2.87) (2.60) (3.01) (5.79) (7.24) Observations 145 145 145 145 145 145 (Pseudo) R2 0.11 0.31 0.25 0.28 0.32 0.46 Note: *p<0.1; **p<0.05; ***p<0.01 Degree of novelty Number of innov. types: 3 types 1.15 2.97 -0.67 1.33 2.85 -0.20 95% Confidence interval for OLS estimate 95% Confidence interval for quantile regression estimate Results discussion: The coefficient of degree of novelty at the 90th (top) percentile is positive, statistically significant and greater than the coefficient at the 50th (mid) percentile. For the average top performing firm, every percent increase in the degree of novelty of their innovations contributes to a 7.76 increase in price-tobook value, a significant jump from the negative contribution of -0.84 to priceto- book value for the average mid performer. As seen in the chart, in the top percentile, the coefficient estimate is significantly greater than estimations from OLS model (estimate and confidence interval in orange), providing confidence that the coefficients are indeed different. A similar situation is seen for the innovation types variable. The effect of introducing three types of innovation (i.e. product, process and business model) among top performers is positive, statistically significant and greater than the effect seen among the mid performers. When the average top performer introduce the three types of innovation, the price to book value is like to increase by 4.12 percent, compared to an increase of 0.07 for the average mid performer. As per the chart, in the top percentile, the coefficient estimate is significantly greater than estimations from OLS model. 18 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland 1.08 4.49 -2.34 Quantile regression results ❷ Processes Innovation processes: Quantile regression results for 10th, 25th, 50th, 75th and 90th quantiles of the dependent variable (Standard errors of coefficients in parentheses. Obtained using 200 bootstrap replications) Price-to-book value Quantile regression OLS 10th 25th 50th 75th 90th Degree of process formalisation 1.14 0.81 0.44 0.05 -2.17 -0.08 (1.68) (0.92) (0.97) (1.28) (3.13) (4.82) Breadth of ext. knowledge search -1.99 0.07 0.55 -0.91 -0.54 -7.06 (1.94) (1.07) (1.27) (1.84) (3.44) (6.26) Breadth of collaboration 2.03* 0.65 0.58 1.05 1.11 6.33** (1.04) (0.62) (0.78) (0.83) (2.16) (3.08) Portfolio management. strategy -0.37 -0.50* -0.21 -0.11 -0.64 0.41 (0.53) (0.29) (0.31) (0.43) (1.00) (1.55) Degree of in-house innov. activities 1.08 -0.17 0.02 1.18 -0.21 5.26 (2.07) (1.10) (1.22) (1.83) (4.56) (6.53) Control variables: Company size (log total assets) 0.01 0.15** 0.06 -0.04 -0.01 0.01 (0.11) (0.06) (0.07) (0.10) (0.24) (0.35) GICS sector (dummies) Included Included Included Included Included Included Constant 1.4 -3.22 -1.18 1.18 6.46 1.81 (2.98) (2.18) (2.34) (2.77) (4.47) (7.29) Observations 190 190 190 190 190 190 (Pseudo) R2 0.11 0.12 0.06 0.08 0.09 0.25 Note: *p<0.1; **p<0.05; ***p<0.01 2.03 3.74 0.32 Breadth of collaboration Degree of in-house innov. activities 95% Confidence interval for OLS estimate 95% Confidence interval for quantile regression estimate Results discussion: The coefficient of breadth of collaboration at the 90th (top) percentile is positive, statistically significant and greater than the coefficient at the 50th (mid) percentile. For the average top performing firm, every percent increase in the breadth of collaboration on innovation activities contributes to a 6.33 increase in price-to-book value, a considerable jump from the contribution of 1.05 to price-to-book value for the average mid performer. As seen in the chart, in the top percentile, the coefficient estimate is significantly greater than estimations from OLS model (estimate and confidence interval in orange), providing confidence that the coefficients are indeed different. A similar situation is seen for the degree of in-house innovation activities. Its coefficient at the top percentile is positive and greater than the coefficient at the mid percentile, although not statistically significant. In addition, the chart indicates that the coefficient estimate is barely different from the OLS estimations. Thus, the contribution of this variable to performance at the top percentile is not as evident as the contribution from breadth of collaboration. 19 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland Degree of entrep. strategic posture Use of cross-functional interfaces Quantile regression results ❸ Attitudes Innovation attitudes: Quantile regression results for 10th, 25th, 50th, 75th and 90th quantiles of the dependent variable (Standard errors of coefficients in parentheses. Obtained using 200 bootstrap replications) Price-to-book value quantile regression OLS 10th 25th 50th 75th 90th Strength of IP protection mech. -1.12 0.91 0.17 -0.85 0.77 -2.65 (1.44) (0.76) (0.72) (1.24) (2.60) (3.42) Use of cross-functional interfaces 0.86 -0.97 -0.76 -0.04 1.24 3.06 (1.75) (1.55) (1.28) (1.29) (3.25) (4.98) Use of network structures -0.03 -0.95 0.12 0.03 0.24 -1.24 (1.77) (1.05) (0.87) (1.50) (2.22) (3.68) Degree of entrep. strategic posture 4.15** 0.03 0.24 1.61 3.29 10.47** (1.67) (1.22) (1.18) (1.45) (3.76) (4.97) Control variables: Company size (log total assets) 0.01 0.14*** 0.08 -0.001 -0.06 0.06 (0.10) (0.04) (0.06) (0.09) (0.19) (0.28) GICS sector (dummies) Included Included Included Included Included Included Constant -0.51 -1.82 -0.99 0.72 2.42 -3.18 (3.06) (1.92) (2.00) (2.49) (4.15) (6.10) Observations 196 196 196 196 196 196 (Pseudo) R2 0.12 0.09 0.04 0.06 0.1 0.26 Note: *p<0.1; **p<0.05; ***p<0.01 6.92 4.15 1.39 0.86 3.75 -2.03 95% Confidence interval for OLS estimate 95% Confidence interval for quantile regression estimate Results discussion: The coefficient of degree of entrepreneurial strategic posture at the 90th (top) percentile is positive, statistically significant and greater than the coefficient at the 50th (mid) percentile. For the average top performing firm, every percent increase in the degree of entrepreneurial strategic posture contributes to a 10.47 increase in price-to-book value, a considerable jump from the contribution of 1.61 to price-to-book value for the average mid performer. As seen in the chart, in the top percentile, the coefficient estimate is significantly greater than estimations from OLS model (estimate and confidence interval in orange), providing confidence that the coefficients are indeed different. A similar situation is seen for the use of cross-functional interfaces when innovating. Its coefficient at the top percentile is positive and greater than the coefficient at the mid percentile, although not statistically significant. Also, the chart indicates that the coefficient estimate is not significantly different from the OLS estimations. Thus, the contribution of this variable to performance at the top percentile is not as evident as the contribution from degree of entrepreneurial strategic posture. 20 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland Robustness tests and post-hoc analyses Analysis of variance: Assess significance of differences across performance quantiles Model variation: Unwinsorised P/B value as alternative DV to assess the effect of winsorisation Model variation: M/B ratio as alternative DV to assess the effect of valuation metric used Model variation: ROA as alternative DV to assess the effect of perf. measure used Model variation: Innovation performance as additional variable to assess the reliability of the estimations We ran ANOVA tests between the We developed supplementary We also developed We also ran supplementary models We added innovation coefficients estimated for the 90th and models with the dependent supplementary models with with ROA (profitability-based performance to our models, as 50th percentile to assess the extent to variable (price-to-book value) market to book ratio (an performance measure) as the it is a potential mediator of the which the coefficients are statistically in its original form alternative market-based dependent variable. Though the innovation-financial significantly different across the (unwinsorised). The effect measure) as the dependent effect sizes and significance of key performance relationship. For performance groups. The test for all sizes, signs and significance of variable. The effect sizes, signs variables differed (as expected), the the four key variables, the signs the key variables indicated that their the key variables remain and significance of the key coefficient sizes for the 90th and effect differences between coefficients at each percentile are similar, confirming that our variables remain similar, percentile remained greater (and 50th and 90th percentiles significantly different at the p < 0.05 choice of winsorization did confirming that our choice of positive) than the 50th percentile. remained similar, confirming level. This supports our insights on the not influence the results market-based indicator did not This confirms our results on the that innovation performance importance of key innovation influence the results. importance of the key innovation does not cause endogeneity variables for top performers compared variables for top performers issues biasing our estimations. to the average. compared to average performers. 21 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland References 1. Manyika J, Chui M, Bughin J. Disruptive technologies: Advances that will transform life, business, and the global economy. McKinsey Global Institute; 2013. 2. Panetta K. Gartner Top 10 strategic technology trends for 2019. Gartner; 2018. 3. EY-Mint. Emerging technologies report 2019: Changing how we live, work and play. EY-Mint; 2019. 4. OECD. Oslo Manual 2018: Guidelines for collecting, reporting and using data on innovation. OECD; 2019. 5. D’Este P, Iammarino S, Savona M, von Tunzelmann N. What hampers innovation? Revealed barriers versus deterring barriers. Research Policy. 2012;41(2):482-8. 6. Laursen K, Salter AJ. The paradox of openness: Appropriability, external search and collaboration. Research Policy. 2014;43(5):867-78. 7. Klingebiel R, Rammer C. Resource allocation strategy for innovation portfolio management. Strategic Management Journal. 2014;35(2):246-68. 8. Cassiman B, Veugelers R. In search of complementarity in innovation strategy: Internal R&D and external knowledge acquisition. Management Science. 2006;52(1):68-82. 9. Teece DJ. Profiting from technological inno-vation: Implications for integration, collaboration, licensing and public policy Research Policy. 1986;15:285-305. 10. Levin RC, Klevorick AK, Nelson RR, Winter SG, Gilbert R, Griliches Z. Appropriating the returns from industrial research and development. Brookings Papers on Economic Activity. 1987;1987(3):783-831. 11. Jansen JJ, Tempelaar MP, Van den Bosch FA, Volberda HW. Structural differentiation and ambidexterity: The mediating role of integration mechanisms. Organization Science. 2009;20(4):797-811. 12. Ordanini A, Maglio PP. Market orientation, internal process, and external network: A qualitative comparative analysis of key decisional alternatives in the new service development. Decision Sciences. 2009;40(3):601-25. 22 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland References 13. Covin JG, Slevin DP. Strategic management of small firms in hostile and benign environments. Strategic Management Journal. 1989;10(1):75-87. 14. Cho HJ, Pucik V. Relationship between innovativeness, quality, growth, profitability, and market value. Strategic Management Journal. 2005;26(6):555-75. 15. Coad A, Rao R. Innovation and market value: a quantile regression analysis. Economics Bulletin. 2006;15(13). 16. Almeida P, Phene A. Subsidiaries and knowledge creation: The influence of the MNC and host country on innovation. Strategic Management Journal. 2004;25(8-9):847-64. 17. Shorack GR. Linear rank statistics, finite sampling, permutation tests and winsorizing. The Annals of Statistics. 1996;24(3):1371-85. 18. Sinha RK, Noble CH. The adoption of radical manufacturing technologies and firm survival. Strategic Management Journal. 2008;29(9):943-62. 19. Eggert A, Thiesbrummel C, Deutscher C. Heading for new shores: Do service and hybrid innovations outperform product innovations in industrial companies? Industrial Marketing Management. 2015;45:173-83. 20. Vega-Jurado J, Gutiérrez-Gracia A, Fernández-de-Lucio I, Manjarrés-Henríquez L. The effect of external and internal factors on firms’ product innovation. Research Policy. 2008;37(4):616-32. 21. Salazar M, Holbrook A. A debate on innovation surveys. Science and Public Policy. 2004;31(4):254-66. 22. Carpenter MA. The implications of strategy and social context for the relationship between top management team heterogeneity and firm performance. Strategic Management Journal. 2002;23(3):275-84. 23. Schmid T, Achleitner A-K, Ampenberger M, Kaserer C. Family firms and R&D behavior–New evidence from a large-scale survey. Research Policy. 2014;43(1):233-44. 24. Murray AI. Top management group heterogeneity and firm performance. Strategic Management Journal. 1989;10(S1):125-41. 25. Coad A, Rao R. Innovation and firm growth in high-tech sectors: A quantile regression approach. Research Policy. 2008;37(4):633-48. 26. Coad A, Segarra A, Teruel M. Innovation and firm growth: Does firm age play a role? Research Policy. 2016;45(2):387-400. 27. Koenker R, Hallock KF. Quantile regression. Journal of Economic Perspectives. 2001;15(4):143-56. 23 | Thriving through innovation: Lessons from the top, Methodological appendix | A timely national study by CSIRO and The University of Queensland CSIRO Australia’s National Science Agency