Research Paper

3.pdf

Home >Business & Finance homework help >Research Paper

What’s driving innovation in small businesses in Australia? The case of the food industry*

Franklin A. Soriano, Renato A. Villano, Euan M. Fleming and George E. Battese

†

There is strong evidence that innovation is a primary driver of a nation’s economic growth. As Australia continues to compete in the global economy, it is imperative that businesses should be innovative to improve their performance. In this paper, we evaluate the status and main drivers of innovation in small businesses in the food sector in Australia. Discrete choice modelling and bootstrapping procedures are applied to a panel of firm-level data collected through the ABS Business Character- istics Survey (2006–2007 to 2010–2011 for the Australian Bureau of Statistics’ Business Longitudinal Database Confidential Unit Record File) to investigate the factors affecting the likelihood of small food businesses to innovate. Results show that businesses are more likely to innovate if they collaborate, have higher information and communication technology intensity, and use science, technology, engineering and mathematics skills. We also found that small food businesses, even at the subsector level, do combine different types of innovation when innovating. The propensity to innovate also increases for small businesses that have flexible working arrangements, face moderate-to-strong market competition, operate overseas and seek finance through debt and equity. The relative importance of these factors was found to vary between agricultural and nonagricultural food subsectors.

Key words: Australia, collaboration, science technology engineering and mathematics skills, information and communication technology, innovation.

1. Introduction

The Australian Government is strongly committed to developing an internationally competitive and productive environment for small businesses. Small businesses

1 play a significant role in the Australian economy,

* Earlier version of this paper was presented at the 61st AARES Annual Conference, Brisbane, 8–10 February 2017. Comments and suggestions from the participants are hereby acknowledged. The assistance of the ABS Customised and Microdata Delivery Unit in the access of the ABS Business Longitudinal Database Confidential Unit Record File (ABS BLD CURF) through ABS Data Laboratory is also acknowledged.

† Franklin A. Soriano, PhD Candidate (e-mail: fsorian2@myune.edu.au), Renato A. Villano,

Professor, Euan M. Fleming, Professor and George E. Battese, Professor are with the UNE Business School, University of New England, Armidale, New South Wales 2350, Australia.

1 The concept of a small business is quite intuitive, and there is no consistently used

definition. The OECD defines it as a firm with generally fewer than 50 employees (OECD 2005b). This study employs the ABS definition of an actively trading business with less than 20 employees. Categories of businesses include the following: nonemploying business–sole proprietorships without employees; microbusinesses (businesses employing fewer than five people), including nonemploying businesses; and small businesses (businesses employing five or more, but fewer than 20, people) (ABS 2002).

Australian Journal of Agricultural and Resource Economics, 63, pp. 39–71

The Australian Journal of

Journal of the Australian Agricultural and Resource Economics Society

mailto:

accounting for almost half of the employment in the private nonfinancial sector and over one-third of production (Australian Bureau of Statistics (ABS) 2015b). They account for the majority of the Australian business counts (97.5 per cent) in 2016–2017 (ABS 2018) and are believed to be the critical players in Australia’s economy, underpinning growth and innovation, and providing jobs for millions of Australians (Australian Government 2015). A recent study by Hendrickson et al. (2015) established that young small and medium enterprises (SMEs) made the highest contribution (40 per cent) to net job creation in Australia over the period 2006 to 2011. Along with these small businesses, the Australian Government considers

the food and agribusiness industry an important part of the Australian economy because it makes significant contributions to the economies of regional areas in employment, business and service opportunities (Depart- ment of Industry, Innovation and Science (DIIS) 2017). Australia’s food and agribusiness industry production was around A$ 54 billion of industry value added in 2014–2015 which is equivalent to 3.3 per cent of the total GDP. Its labour productivity has increased at an annual rate of 2.4 per cent over the past five years since 2010. The industry’s exports were worth A$ 40.8 billion in 2015 (Chaustowski and Dolman 2016). With this industry spending nearly A$ 728 million on research and development (R&D) related to food, beverage and agricultural machinery manufacturing, an important question is as follows: Have the small food businesses been engaging in any form of innovation activities that are useful in improving their performance? Innovation remains vital to the expansion and international competitive-

ness of Australia’s economy and, as Australia continues to compete in the global economy, Australian businesses such as the small food businesses need to be innovative for job creation and income growth (DIISR 2009, 2018a). Innovation statistics in selected growth sectors

2 in Australia show that only

32.7 per cent of all businesses in food and agribusiness were engaged in any innovative activity in 2013–2014 of which 28.2 per cent successfully introduced and implemented innovation (ABS 2015d). This was lower than the overall proportion of business innovators in Australia, which was 48.3 per cent in that period. In addition, the innovation rate for Australian small businesses employing 5–19 persons was 59.9 per cent in 2013–2014 (ABS 2015c). Evidently, there is a need to encourage small businesses in the food and agribusiness industry to innovate to become more competitive and productive. To unleash the potential of small food businesses to innovate, grow and create more jobs, it is imperative to examine more closely what drives them. This is what the current study seeks to achieve. Central to the vision of providing the right economic incentives to enable

businesses – big and small – to grow, the Australian Government through the

2 The food and agribusiness growth sector includes food-related production, food

processing and the major inputs into these sectors, but not the wholesale and retail sale of these goods (ABS 2015d).

40 F. A. Soriano et al.

Industry Innovation and Competitiveness Agenda3 has identified food and agribusiness as an area of competitive strength and prioritised it as a growth sector (Department of the Prime Minister and Cabinet 2014). A key initiative in the agenda was establishing Industry Growth Centres and, in the case of the Australian food and agribusiness sector, the Food and Agribusiness Growth Centre, known as Food Innovation Australia Ltd (FIAL), which was established to build capability and encourage collaboration and innovation in the industry involved (DIIS 2017). For the growth centre’s plans, activities or strategies to drive innovation, productivity and competitiveness and help food and agribusiness businesses build stronger futures for themselves, and it is essential to provide empirical evidence of factors that drive these businesses to innovate. Empirical studies on the determinants of innovation in the Australian food sector are nonexistent, so a study on small food businesses in Australia using firm-level data is desirable to support the above-mentioned agenda. In the light of the above discussion, this study aims to evaluate the status

and main drivers of innovation in small businesses in the food industry in Australia using a discrete choice model with panel data. In particular, we examine how large the impacts of these key drivers are on business innovation, and what are the implications of the findings for the govern- ment’s innovation agenda and initiatives. What is compelling in this work is the use of the food industry sample in the five waves, (2006–2007 to 2010– 2011, of the ABS Business Longitudinal Database (BLD)), a unique extract of longitudinal microdata that have never been used by any previous researcher to analyse business innovation or performance in small businesses in the Australian food industry using panel data. The data also allowed us to go deeper and analyse the food industry subsectors. Another remarkable contribution of our study is the estimation of robust standard errors for the impact measures called ‘average partial effects’ via simulation using modern bootstrapping procedures as well as analysing their distribution. This empirical study also adds to the literature on the analytical use of the ABS confidentialised business survey microdata. The paper proceeds as follows. Section 2 presents a review of the relevant

literature which formed the basis for the hypotheses and conceptual framework of this study. Section 3 describes the data and the methodology used in the empirical analysis. Sections 4 and 5 provide the findings and conclusions.

2. Literature review and framework

This section examines the theoretical and empirical literature on innovation that describes the key determinants and relationships examined in this study.

3 More information regarding the Industry Innovation and Competitiveness Agenda and

Growth Centres can be found at the DIIS website: www.industry.gov.au.

What’s driving innovation in small food businesses in Australia? 41

http://www.industry.gov.au

The choice of the key determinants is limited to the information being collected in the ABS Business Characteristics Survey (BCS) and what has been used in the previous ABS cross-sectional studies on innovation. Particular attention is given to the food industry studies on innovation by other countries and those previous ABS and DIIS innovation analyses (mostly cross-sectional). Included also are the conceptual framework, some definitions and the hypotheses to be tested.

2.1 Defining innovation

This study uses the Organisation for Economic Co-operation and Develop- ment (OECD) definition of innovation as provided in the Oslo Manual:

. . .the implementation of a new and significantly improved product (good or service); or process, a new marketing method, or a new organisational method in business practices, workplace organisation or external relations.(OECD 2005a, p. 46)

A business is called ‘innovation-active’ if it engages in any innovation activities that were implemented, ongoing or abandoned during a period. A business is called an ‘innovator’ if it successfully developed and implemented an innovation, which may have taken many years to complete (OECD 2005a). The current investigation is conducted for ‘innovation-active’ businesses. The study uses an output-based measure of innovation obtained from the

ABS BCS data. Four types of innovation are scoped in the analysis: product innovation, organisational or managerial innovation, operational process innovation and marketing innovation (OECD 2005a). Product innovation is any good or service or combination of these which is new (or significantly improved) to a business. New or significantly improved strategies, structures or routines of a business that aim to improve performance belong to organisational or managerial innovations. Operational process innovations refer to new methods of producing or delivering the products including significant changes in techniques, equipment and/or software. And, new design, packaging or sales methods to increase the appeal of goods or services of a business or to enter new markets are classified as marketing innovations. Of course, a business could undertake more than one type of innovation. The earliest work on innovation came from Schumpeter (1942), who

described capitalism as a form or method of economic change which can never be stationary. The fundamental impulse that sets and keeps the capitalist engine in motion comes from new consumer goods, new methods of production or transportation, new markets or new forms of industrial organisations those capitalist enterprises create. Schumpeter developed a theory where an enterprise’s ability to innovate was mainly due to its size, stating that large businesses are the main drivers of innovation, being better resourced and having monopolistic power in the market. In this study, we

42 F. A. Soriano et al.

provide supportive evidence that small businesses in the Australian food sector can be strong innovators too.

2.2 Innovation and collaboration

According to the Oslo manual, the innovative activities of a firm partly depend on the variety of linkages to sources of information, knowledge, technologies, practice, and human and financial resources. One of the external links is innovation cooperation, which requires active collaboration with other firms or research institutions on innovation activities (OECD 2005a). The Nelson and Winter (1982) model states that the organised R&D efforts of firms are the sources of innovation. This model views innovation as a path-dependent process where knowledge and technology are established via interaction between various participants (firms, institutions and society) known as the evolutionary approach. Closely intertwining this are the systems of innovation approaches (Nelson 1993 and Lundvall 2010) where external institutions influence the innovation activities of firms through the transfer of ideas, skills, knowledge and information. The above models link collaboration with innovation. Another model receiving much attention recently that works very well with

collaboration is open innovation (OI). Chesbrough (2003) introduced and defined OI as the use of purposive inflows and outflows of knowledge to accelerate internal innovation, and expand the markets for external use of innovation. The OI concept has been applied in several studies. For example, Bigliardi and Galati (2016) indicated that SMEs operating in less innovative industries give emphasis to collaboration as barriers to OI. Henttonen and Lehtim€aki (2017) show that technology-intensive SMEs engaged in OI for commercialisation rather than for R&D. Marangos and Warren (2017) examined what strategies the senior managers of R&D-intensive SMEs in the life sciences sector conduct concerning OI. Studies by Bigliardi and Galati (2016), Henttonen and Lehtim€aki (2017), and Marangos and Warren (2017) also provide good literature reviews on OI in SMEs. An extensive review of the literature regarding the adoption of OI in the food industry is by Bigliardi and Galati (2013). The adoption of OI practices in the food industry is still emerging, and requires attention and further investigations from academia using firm-level data. Analysing the relationship between collaboration and innovation in the

food industry is not new in the international arena. There is evidence that innovation is influenced when businesses collaborate externally. De Martino and Magnotti (2017) found that partnerships between academia and research institutions were crucial in boosting the innovative capacity of small food firms in Italy. Galati et al. (2016) looked at how networking with innovation partners is needed for food firms to perform effective OI. Innovation in Italian food businesses has also been shown to rely on R&D activities and knowledge sourcing with external partners (e.g. universities, customers,

What’s driving innovation in small food businesses in Australia? 43

suppliers, competitors, public or private institutions) (Maietta 2015; Ciliberti et al. 2016a,b). Wixe et al. (2017) and Caiazza and Stanton (2016) found that innovation in small food firms had significant association with external collaboration to access knowledge. A study by Grunert et al. (1997) in Denmark found that the interaction among R&D, market orientation and collaboration are major determinants of innovation in the food sector. McAdam et al. (2014) observed that horizontal collaboration (i.e. between firms in the same industry, not competing directly but selling and marketing to similar customers) among SMEs in the UK agri-food sector also played a significant role. However, in Australia, studies on collaboration and innovation in the food industry are minimal and only focused on the descriptive profiling of industries without quantitative measures with respect to the determinants of innovation (ABS 2015d). Collaboration is highly valued in Australian innovation systems. It is

defined as an interactive process that involves more than two individuals, businesses or organisations working cooperatively towards a common goal, including some sharing of technical and commercial risk (DIIS 2008). Innovation and collaboration are high priorities for the Australian Govern- ment because working with researchers can give Australian businesses a competitive edge. The Cooperative Research Centres Program is a good example of the initiative led by the DIIS focusing on collaborative research partnerships between industry entities and research organisations (DIIS 2018b). In relation to the food industry, FIAL recently released Australia’s first food, beverage and agribusiness cluster initiative that would benefit farmers, food suppliers and retailers across Australia. This program encourages businesses, researchers and educational institutions to work together to build on their comparative advantage and develop innovations (FIAL 2017). For policymakers like the DIIS, understanding if collaboration enables, or is at least associated with, innovation among small food industry businesses is of great interest. In this study, collaboration is measured as a binary response when a

business is involved in any of the following collaborative arrangements: joint R&D, joint buying, joint production of goods and services, integrated supply chain, joint marketing or distribution, and other collaborative arrangements specified by the business. It can be simple partnerships between suppliers and customers, linked networks on common interests, associations of like industries, industry or business groups, research institutes, joint ventures and entities. It has been established in some ABS and DIIS research studies that collaboration plays a significant role in driving business innovation. It is found to be strongly and positively associated with a higher likelihood of being an innovator (DITR 2006; ABS 2008; Rotaru 2013; Rotaru and Soriano 2013). To provide empirical evidence to support the proposition that this relationship also holds true in the case of the small food businesses in Australia, we seek to test the hypothesis: ‘Small food industry firms that collaborate are more likely to innovate’ (Hypothesis 1).

44 F. A. Soriano et al.

2.3 Innovation and skills

Much innovation knowledge is embodied in people and their skills, and appropriate skills are needed to make use of external sources of innovation (i.e. the role of human capital in a firm’s engagement in any form of innovation); for example, how well the skills of employees match the needs of innovative firms and collaborators (OECD 2005a). According to OI strategy, firms need smart and technologically capable managers to properly commu- nicate and exchange ideas with external partners (Marangos and Warren 2017). Galati et al. (2016) studied how personnel skills were needed for food firms to perform effective OI. The link between skills and innovation is supported by the human capital

theory and resource-based theory. Becker (1994) viewed human capital as directly useful in the production process (and the process connects to innovation). More explicitly, human capital increases the productivity of workers in different tasks. The Australian Workforce and Productivity Agency (2013) compiled international and national literature that establishes skills that play a key role in productivity. The resource-based theory (Grant 1991) established that both resources and capabilities of a firm (hereby human capital) are central in formulating its strategy that provides direction as well as the primary source of profit. Resources include the skills of employees which are inputs in the production process, and capabilities are the capacity of employees to perform tasks. Toner (2011) reviewed the literature on the role of workforce skills in the

innovation process (to incremental innovation, in particular) in developed economies. It draws from the innovation studies discipline, neoclassical human capital theory, institutionalist labour market studies and the work organisation discipline. It showed that the quantity and quality of workforce skills are a major factor in determining the observed patterns of innovation and are key aspects of economic performance. The quality of human capital, as an important driver of process and product innovation, has also been seen in the works of Capitanio et al. (2009, 2010). In the food sector, Huiban and Bouhsina (1998), Avermaete et al. (2004), Smit et al. (2015), Vancauteren (2016), and Brown and Roper (2017) showed that skills of the workforce play a key role in business innovation. A recent ABS study by Soriano and Abello (2015) found that the

probability of innovation is higher for a business that uses employees with science, technology, engineering and mathematics (STEM) skills than for a business that does not use these skills. The impact of the use of STEM skills on the probability of having ‘new-to-the-world’ type of innovation is also found to be strong and positive. Following the work of Soriano and Abello (2015), skills in this study are assessed using the STEM qualifications of employees used by the firm. STEM skills are defined to be present in a firm, according to the Australian Standard Classification of Education (ABS 2001), if employees have qualifications of one or more of the following:

What’s driving innovation in small food businesses in Australia? 45

postgraduate degree, master degree, graduate diploma, graduate certificate, bachelor degree, advanced diploma and Certificates II and IV, in any of the fields of the natural and physical sciences (including mathematical sciences); information technology (IT); engineering and related technologies; and agricultural, environmental and related studies. The STEM skill variables are constructed based on the type of skills used by a business, as reported in the ABS BCS. A business is considered to have used STEM skills if it reported using any of the following skills: engineering, scientific research, IT professionals and IT support technicians. These are based on subjective responses by businesses to the BCS question about the types of skills used in undertaking core business activities. It is worth noting that a particular business may use any of the above STEM skills in combination with other non-STEM skills such as trade, transport, and plant and machinery operation. To determine whether skills of employees drive the small food businesses in Australia to innovate, we want to test the hypothesis ‘Small food industry firms that use STEM skills are more likely to innovate’ (Hypothesis 2).

2.4 Innovation and ICT

A conceptual link between the adoption of information and communication technology (ICT) and innovation was established by K€ollinger (2005) who argued that ICT has strategic relevance for firms because it is a valuable source of business innovation providing substantial efficiency gains. ICT technologies that create automated system links lead to more streamlined businesses processes and enable employees to develop closer links between businesses, their suppliers, customers, competitors and collaborative partners, allowing the business to be more responsive to innovation opportunities (Todhunter and Abello 2011; Domenech et al. 2014; Salim et al. 2015; Galati et al. 2016). The same arguments were discussed and analysed by Gretton et al. (2003) and Tiy et al. (2013) using Australian data, indicating that ICT plays an important role in business innovation. Gago and Rubalcaba (2007) found that businesses that invest in ICT, particularly those that regard their investment as strategically important, are significantly more likely to engage in services innovation. In regard to the food industry, Galati et al. (2016) revealed that ICT was needed for Italian food firms to perform effective OI. OECD (2003, pp. 3–4) defined ICT as ‘goods . . . that are either intended to

fulfil the function of information processing and communication by electronic means, including transmission and display, or which use electronic processing to detect, measure and/or record physical phenomena, or to control a physical process’. In this study, we combine broadband Internet connection, business Web presence and use of e-commerce into an ICT intensity index to develop a convenient and meaningful measure of ICT sophistication (i.e. business not having broadband connection (low ICT intensity) to having the three components of innovation (most intense ICT). This index was also

46 F. A. Soriano et al.

applied by Todhunter and Abello (2011) and Tiy et al. (2013). Using the same index, econometric analyses at ABS have shown that there is a positive and significant association between business innovation and the use of ICT (Todhunter and Abello 2011; Rotaru 2013; Rotaru and Soriano 2013; Rotaru et al. 2013; Tiy et al. 2013). These studies reveal that more intense ICT users are likely to undertake more types of innovation. The same hypothesis is tested here for small food businesses in Australia, stated as ‘Small food industry firms that have higher ICT intensity are more likely to innovate’ (Hypothesis 3).

2.5 Innovation and flexible working arrangements

Is labour market flexibility good for innovation? Can businesses gain and maintain a competitive edge in the global and ever-changing and uncertain environment if they allow this flexibility? Labour market flexibility, being a dynamic concept, has received considerable and growing attention in recent years (Beatson 1995), in which ‘flexibility’ was defined as the ability of markets (and the agents that operate in them) to respond to changing economic conditions. There are different forms of labour market flexibility and, following Atkinson (1984), Beatson (1995), Michie and Sheehan-Quinn (2001), Reilly (2001) and Kleinknecht et al. (2006), labour market flexibility can be categorised into five types:

• Numerical flexibility is the ability of firms to allow variation in the number of employees or workers used. Examples are temporary, seasonal, casual, outsourcing and fixed-term workers.

• Temporal flexibility is the ability of firms to allow variability of working hours. Examples are part-time, shift, reduced hours, overtime and leave flexibility.

• Functional flexibility is the ability of firms to allocate their labour force to carry out a wide range of task and activities. Examples are multiskilling, task flexibility and cross-functional working.

• Locational flexibility is the ability of firms to use employees outside the normal workplace. Examples are home-based work, use of mobile phones and tele/outworkers.

• Financial or wage flexibility is the ability of firms to decide wage levels in line with corporate performance. Examples are gain or profit sharing, wage- cutting deals and pay increases based on performance.

Studies that examined labour market flexibility (following any of the above types) and innovation have been increasing (Michie and Sheehan-Quinn 2001; Storey et al. 2002; Michie and Sheehan 2003; Kleinknecht et al. 2006, 2014; Zhou et al. 2011). Kleinknecht et al. (2006) followed Schmooklerian demand-pull theory (see Schmookler 1966) where higher effective demand increases innovative activity and labour productivity. In relation to labour

What’s driving innovation in small food businesses in Australia? 47

market flexibility, they found that wage restraint or downward wage flexibility impedes innovation. In terms of numerical flexibility, the study noted that a large inflow of new people (skilled and productive) enriched the pool of innovative ideas within firms and led to new collaborations. Zhou et al. (2011) found that firms with high shares of workers on fixed-term contracts tend to have fewer sales of innovative new products, while high functional flexibility enhances firm’s new products sales. Chung (2009) observed that temporal flexibility had an effect on innovation, whereas Martinez-Sanchez et al. (2008) revealed that innovation performance was positively associated with internal functional flexibility and negatively associated with numerical flexibility and outsourcing, but the relationships between them were moderated by interorganisational cooperation. Rotaru (2013) and Rotaru and Soriano (2013) were the first to examine the

relationships between four types of flexible working arrangements (flexible working hours, flexible leave, job sharing and working from home) and innovation using Australian business survey data at the firm level. Both studies found that all these four factors played important roles in influencing innovation. The degree of relationship varied but was mostly positive and significant. We know of no study on labour market flexibility and innovation in the food industry; hence, this paper is a pioneering work that establishes the role of work flexibility on innovation for small food businesses in Australia. Thus, we hypothesise that ‘Small food industry firms that have flexible working arrangements are more likely to innovate’ (Hypothesis 4). The variable for flexible working arrangements in this paper is limited to a

business offering their employees any of the above four arrangements, as in Rotaru (2013) and Rotaru and Soriano (2013).

2.6 Innovation and competition

Studying the relationship between market competition and innovation has long been of interest to economic researchers and policy analysts. Schum- peterian theory states that greater levels of market competition faced by businesses lead to declining innovation, but most empirical works have found the opposite result. There is a wealth of theoretical literature on the connection between competition and innovation, but a more recent theory by Aghion et al. (2005) suggests that the relationship between the two has an inverted-U shape. This framework is an attractive one because it reconciles the Schumpeterian and non-Schumpeterian results (i.e. increased competition is associated with more innovation). In addition, the industrial organisation theory of Tirole (1988) also emphasised the importance of competitive positioning – that businesses innovate to remain in their competitive position and avoid losing (see Correa (2012) for a similar framework to Aghion et al. (2005)). There is a wealth of international empirical literature that has examined the

link between competition and innovation using firm-level data, but examples

48 F. A. Soriano et al.

are few in Australia (Bhattacharya and Bloch 2004; Rogers 2004; Wong et al. 2007; ABS 2008; Griffiths and Webster 2009; Soames et al. 2011; Rotaru and Soriano 2013; Rotaru et al. 2013; Soriano and Abello 2015; Palangkaraya et al. 2016). Soames et al. (2011) obtained empirical results that supported the Aghion et al. hypothesis. Similar results were reported by Rotaru et al. (2013), Rotaru and Soriano (2013), and Soriano and Abello (2015) using the ABS BCS data. In this paper, the degree of market competition is divided into four

categories: no effective competition (no competitors); minimal competition (1–2 competitors); moderate competition (3–4 competitors) and strong competition (five or more competitors). We are interested to test whether the theory of Aghion et al. (2005) holds in the case of small food businesses in Australia. This will add to the empirical literature on the link between competition and innovation in the food industry, which is minimal at the moment. The specific hypothesis is ‘Small food industry firms that face moderate-to-strong competition are more likely to innovate’ (Hypothesis 5).

2.7 Other drivers of innovation from food industry studies

Empirical studies using firm-level data in the food industry have also observed some other important determinants of innovation, one being exporting behaviour of firms. The modern trade and growth theories by Grossman and Helpman (1991), and Aghion and Howitt (2009) suggest that access to export markets affects innovation. The recent work of Aghion et al. (2018) provides evidence where a simple model of trade and innovation predicts that a positive export shock increases market size and, therefore, innovation incentives for more productive businesses. A study on the food industry by Domenech et al. (2014) indicated that the adoption of ICT technologies was a key factor for innovation in agri-food industries in Spain, and it was highly influenced by firm size and export orientation. Another study of Spanish agri-food firms showed that export was positively associated with the four types of innovation (Zouaghi and Sanchez 2016). Providing evidence on the relationship between export and innovation in small businesses in the food industry has received little attention, although there is evidence showing that Australian businesses engaging in any export activity are more likely to innovate (Soames et al. 2011; Rotaru 2013; Tuhin 2016; Tuhin and Swanepoel 2017). Another important driver of innovation is having good financial support,

either internally or externally, to fund innovations. Getting access to finance is of crucial importance for investing in innovations (Sauer 2017). In Australia, lack of access to additional funds was the most commonly reported barrier (17 per cent) to undertaking innovation (ABS 2017). Rotaru et al. (2013) applied propensity score matching in the context of causal modelling, using the ABS BCS data, and found a statistically significant association

What’s driving innovation in small food businesses in Australia? 49

between government assistance and innovation. De Martino and Magnotti (2017) also concluded that public funding is crucial in boosting the innovative capacity of small food firms in Italy. Other food industry studies like that by Avermaete et al. (2003) found that

innovation depends on business age and regional economic performance. Zouaghi and Sanchez (2016) found a significant impact of the global financial crisis (GFC) on the innovative performance of agri-food firms. The year effects were considered in this current study to capture the impact of GFC. The authors also found a local study that examined the important role of intellectual property (IP) in the food innovation process. IP Australia (2014) used patent analysis to assess the scope, quality and impact of innovative activity in the food sector and showed that Australia exhibits a positive technological specialisation in the food industry and collaboration is prevalent among universities and research institutes. From the above food industry studies on innovation, it can be gleaned that

other factors such as export, finance, age of firms, patents and an external factor like the GFC may be associated with firms’ propensity to innovate. Smith and Hendrickson (2016) also verified the importance of a firm’s age in their analysis of Australian SMEs. Unfortunately, patents and age informa- tion are not currently available in our data set.

2.8 The conceptual framework

In view of the above literature and hypotheses, the conceptual framework employed in this study is presented in Figure 1. The above framework shows the key determinants for business innovation

that are expected to play an important role in small businesses in the Australian food industry. The determinants (i.e. collaboration, the use of resources with STEM skills, ICT, flexible working arrangements, competi- tiveness, export capability, access to finance and business size) were selected or have been limited by the availability of data on small food businesses in the ABS BCS. These drivers are believed to push businesses in the small food industry in Australia to engage in forms of innovation, thereby providing empirical evidence that will support the Australian Government’s Industry Innovation and Competitiveness Agenda. Establishing the connections between collaboration, the use of STEM skills, ICT and innovation among small businesses in the Australian food industry are also very important for the National Innovation and Science Agenda (Australian Government 2017). Such evidence serves as an input for the DIIS policy evaluation regarding innovation and productivity growth in the food industry. The directions of the arrows in Figure 1 show their interdependency relationships. The general focus of the analyses is for one aspect of innovation – the propensity of businesses to innovate (i.e. overall innovation). However, additional analysis incorporating the four types of innovation using multivariate modelling is undertaken to complement the main findings. The study also examines the

50 F. A. Soriano et al.

effects of the above determinants on the food industry subsectors – the small food businesses belonging to the agriculture, forestry and fishing industry, and manufacturing and wholesale trade industries.

3. Data and methods

3.1 Data source

This study uses firm-level data from the ABS BLD contained in the ABS Business Longitudinal Analysis Data Environment (ABS BLADE).5 The BLD contains longitudinal data from small and medium businesses that are designed to allow analysis of micro drivers of business innovation and performance over time. The BLD is specifically designed for longitudinal analyses and not to produce reliable aggregated/population information/

Goods and services Organisational and managerial Marketing methods

Innovative outputs (Overall innovation)

Collaboration

Use of STEM skills

Use of information and communication technology

Other key business characteristics

• Export capability • Finance • Business size • Subsector where firm belongs

Small “Food

industry” business

G o v e r n m e n t

P o l i c y

Operational process

Flexible working arrangement

Degree of market competition

Figure 1 Conceptual framework for the study.

4 These are the only Australian and New Zealand Standard Industrial Classification

(ANZSIC) division 1 industries covered in the ABS Business Longitudinal Data (BLD) used in this study. Businesses sampled are predominantly associated with food for human consump- tion in the above three industry divisions (ABS 2013).

5 Formerly known as the ABS Expanded Analytical Business Longitudinal Database (ABS

EABLD) (ABS 2015e).

What’s driving innovation in small food businesses in Australia? 51

estimates. Successive panels are followed for five years, with collected characteristics data using the BCS linked to a small number of financial data items sourced from the Business Activity Statement (BAS) data provided to the ABS by the Australian Taxation Office. The BCS collects data on key indicators of business performance, use of ICT and innovation in Australian businesses (ABS 2015a,c). The BLD is made available to researchers via a confidentialised unit record file (CURF) that is accessible through the ABS Data Laboratory.

The empirical analysis specifically utilises the third panel of the ABS BLD CURF (ABS 2013). This BLD CURF comprises one sample (referred to as a panel) that is drawn from the in-scope Australian business population at 30 June 2007. This panel contains data from this one sample over five reference periods (2006–2007, 2007–2008, 2008–2009, 2009–2010 and 2010–2011). The total sample size is 3,075. The BLD sample design is based on the panel that represents the Australian business population when the panel is introduced in the BLD. The scope of the BLD is actively trading businesses with a simple structure in the Australian economy. The requirement to include the food industry component in the first three

panels of the ABS BLD arose in 2003 following the commissioned project of the Department of Agriculture, Fisheries and Forestry to analyse the business performance of small and medium businesses in the Australian food industry. This component was developed by including an additional sample in the three relevant industry divisions: agriculture, forestry and fishing (AgriFF); manufacturing (MFG); and wholesale trade (WT) (non-AgriFF = MFG and WT). It is defined by the 2006 Australian and New Zealand Standard Industrial Classification (ANZSIC06) classes predominantly associated with food for human consumption in the above three industry divisions (ABS 2006). The third BLD CURF panel is the last panel to include the food industry component (ABS 2013). Of the 3,075 businesses sampled for the third BLD Panel, 984 businesses

are flagged belonging to the food industry. Table 1 presents the distribution of the sampled units by business size. The volume of data included in the BLD is substantial, and users should take into consideration cleaning and quality checking them before using them in any analysis. Data may be missing in the BLD CURF due to changes in the survey questions, sampled businesses ceasing operation (either temporary or permanent death), sampled businesses undergoing structural change, survey nonresponse and other reasons. After cleaning and quality checking the food industry sample, particularly the data items required for the panel modelling, the current study ended up with a balanced panel of 417 small businesses, 237 belonging to AgriFF and 180 coming from the combined manufacturing and wholesale

6 For more information on the ABS Data Laboratory, see Parker (2017).

52 F. A. Soriano et al.

trade divisions. Although part of the original BLD, the medium-sized food businesses are not included in our study. The panel analysis discussed in Section 4 has been shortened to four years

(i.e. 2007–2008 to 2010–2011) because the questions on business use of STEM skills and flexible working arrangements were not available in the 2006–2007 BCS questionnaire.

3.2 Methods

This subsection presents the methodological framework employed to address the objectives stated in Section 1 and in testing the five hypotheses outlined in Section 2. We follow Wooldridge (2010) to provide the theoretical under- pinnings for the various empirical models. The main panel modelling approach utilised in our study mostly follows

that of Rotaru (2013), and Rotaru and Soriano (2013). These studies are the first ABS methodological papers that explore the implementation, estimation and performance of different discrete choice longitudinal data models using the Main Unit Record File of the ABS BLD. The models include the following: the pooled model; the standard random-effects model; the correlated random-effects model, using the specifications of Mundlak (1978) and Chamberlain (1984); a standard dynamic model; and a dynamic random-effects probit model that follows Wooldridge (2005). The standard random-effects probit model is employed. It is one of the

most popular and widely used nonlinear models for binary outcomes with panel data (Wooldridge 2005, 2010, 2013; Greene 2012; Hensher et al. 2015). It has been applied also to model innovation propensity in the food industry (Triguero et al. 2013; Vancauteren 2016). A key assumption for this model is that the observed covariates are strictly exogenous, conditional on the unobserved effect.

Table 1 Food industry panel sample size in the ABS BLD CURF panel three, 2006–2007 to 2010–2011

Nonemploying businesses no.

Micro (1–4

persons) no.

Small (5–19

persons) no.

Medium (20–199

persons) no.

All business size groups

No.

BLD sample AgriFF 144 144 132 129 549 Non-AgriFF 135 111 96 93 435 Total 279 255 228 222 984

Sample used in the modelling AgriFF 64 88 85 237 Non-AgriFF 59 56 65 180 Total 123 144 150 417

What’s driving innovation in small food businesses in Australia? 53

For firm i at time t, let xit be a vector of (observed) explanatory variables. The explanatory variables include the following business characteristics: business size; subsector where the business belongs; collaboration status; business use of STEM skills; business use of ICT (or ICT intensity); degree of market competition; business export capa- bility

7 ; flexible working arrangements; business finance (i.e. business seeks

debt and/or equity); time period (year effects); and a constant term. The construction of the variables included in the model follows the ABS studies mentioned in Section 2 (ABS 2008; Todhunter and Abello 2011; Rotaru 2013; Rotaru and Soriano 2013; Rotaru et al. 2013; Tiy et al. 2013; Soriano and Abello 2015). Interested readers are referred to these studies for more details. We consider the dependent variable, yit, as a binary response variable

taking the value 1 if the i-th business innovated in the t-th year, and 0 otherwise. Assume that the values taken by yit are determined by a latent variable, y�it, given by

y�it ¼ x0itb þ ai þ eit; where i ¼ 1; k; n; t ¼ 1; 2; 3; 4; ð1Þ

where y�it is the unobserved binary variable which corresponds to yit, the observed dichotomous variable for firm i at time t. The relationship between yit and y

� it is defined by

yit ¼ 1 if y�it [ 0 0 if y�it � 0

� ð2Þ

xit is a vector of observed covariates including a constant term, as defined earlier; b is a vector of fixed, yet unknown, population parameters; ai stands for the random component (unobserved heterogeneity); and eit is the error term, and ɛit|xit � N(0,1). For a balanced panel, the random-effects probit model has the form:

Pðyit ¼ 1 ��xi; aiÞ ¼ Pðy�it [ 0��xit; aiÞ

¼ Pðx0itb þ ai þ eit [ 0 ��xit; aiÞ

¼ Uðx0itb þ aiÞ ð3Þ

Note that xit appears in the model, although xi = (xi1, xi2, xi3, xi4) is in the conditioning set and where Φ(�) is the normal cumulative distribution function for eit conditional on xi and ai.

7 In this analysis, the export capability variable is defined as business selling products to

local and/or overseas markets. Following the ABS BCS, local (geographic) market includes the immediate area, town or city in which the business is located as well as outside this area but within Australia.

54 F. A. Soriano et al.

It follows that the joint probability function is as follows:

Pðyi1; . . .; yi4 ��xi; aiÞ ¼ Y4

t¼1 P yit

��xit; ai� �: ð4Þ The relationship between yit and ai above is linear and additive in

functional form, and we assume independence between regressors and the unobserved heterogeneity. The conditional distribution of ai (assuming the normal distribution) is

given by ai ��xi � Nð0; r2aÞ; ð5Þ

and ai is independent of the error term eit. The standard random-effects probit model is estimated using the method of maximum likelihood. To measure the relative importance of the unobserved effect, the correlation between the composite latent errors, say, ai + eit, across any two time periods (t and s), is estimated as

Covðvit; visÞ ¼ rhoðqÞ ¼ r2a

r2a þ r2e where vit ¼ ai þ eit; t 6¼ s: ð6Þ

In this case, r2a is the variance of the unobserved effects, r 2 e is the variance

of the idiosyncratic component and r2a þ r2e is the composite error. The significance of the estimate for rho is tested using a likelihood ratio test. If the estimate for rho is not significantly different from zero, then the random- effects probit estimator is favoured over the pooled probit estimator. The bootstrap technique was implemented for robust statistical inferences

in the standard random-effects probit model. All statistical computing was executed using Stata 15 MP software. To check the sensitivity of the standard random-effects probit model

results, we also estimated the pooled probit model which does not directly deal with the unobserved firm-specific effects. Because this model ignores the unobserved heterogeneity, one can make adjustments in the estimation by computing panel-robust standard errors. This model is attractive in that it is simple to implement and interpret and because robust standard errors can be obtained without imposing specific functional forms. The model has also been widely used and provides a good reference for the standard random- effects probit model (Wooldridge 2005, 2010, 2013; Greene 2012; Hensher et al. 2015). To complement the sensitivity analysis, a standard multivariate panel

probit model is also estimated using simulated maximum likelihood (see Cappellari and Jenkins 2003; Wooldridge 2010, 2013; Greene 2012). The same technique was employed by Soames et al. (2011) and Srholec (2016) in their research work on innovation. This is undertaken to further investigate the correlation structure between the four types of innovation and to estimate

What’s driving innovation in small food businesses in Australia? 55

simultaneously the effect of the drivers for all innovation outcomes. For this, we considered the four-equation multivariate latent process given by

y�itð1Þ ¼ x0itð1Þbð1Þ þ eitð1Þ; where i ¼ 1; k; n; t ¼ 1; 2; 3; 4 y�itð2Þ ¼ x0itð2Þbð2Þ þ eitð2Þ y�itð3Þ ¼ x0itð3Þbð3Þ þ eitð3Þ y�itð4Þ ¼ x0itð4Þbð4Þ þ eitð4Þ ð7Þ

where

yitðmÞ ¼ 1 if y�

itðmÞ [ 0 0 if y�

itðmÞ � 0 �

; m ¼ 1; 2; 3; 4; ð8Þ

and eitm, m = 1,2,3,4, are error terms distributed as multivariate normal, each with a mean zero and variance-covariance matrix V, where V has values 1 on diagonal and correlations qkl ¼ qlk; 6¼ k; l; k ¼ 1; 2; 3; 4 as off-diagonal elements. Note that the above model is a simultaneous system of four binary probit equations, where the dependent variable, yit(m), is a binary response variable taking the value 1 if the i-th business innovated a particular type of innovation (i.e. product innovation (m = 1); organisational or managerial innovation (m = 2); operational process innovation (m = 3); marketing innovation (m = 4)) in the t-th year, and 0 otherwise. The modelling conducted here is static in relation to the response variable,

and, as such, it is not possible to establish the existence or direction of causality between the various conditioning (business characteristics) variables and the propensity to innovate. The average partial effects (APEs) are then computed to summarise the marginal effects across the distribution of all observable covariates in the balanced sample. The APEs have the advantage of being comparable across the models investigated in this study, an advantage which is not often preserved with the estimated regression coefficients. The APEs are an alternative way to summarise marginal effects. In our study, the APE of any discrete variable (as are most of our covariates) can be thought of as measuring the discrete increment in probability of the small food business to innovate, averaged over the distribution of the unobserved variable(s), usually done by conditioning on a set of values. The APE for a binary variable, xh, is given by

AP̂Exh ¼ 1

Xn i¼1

U x0itb̂ þ âi ��xh ¼ 1� � � U x0itb̂ þ âi��xh ¼ 0

� �h i" # : ð9Þ

The statistic in Equation (9) is simply the average of the discrete differences in the predicted probabilities. Again, robust standard errors for the APEs are calculated by the bootstrapping method. The same technique has been

56 F. A. Soriano et al.

thoroughly explored in Rotaru (2013). To complement each of the estimated APEs, we also examined its distribution focusing on the quantiles (partic- ularly the 25th and 75th percentiles, largest and smallest values), variance, skewness and kurtosis.

4. Findings

This section presents the empirical results of the econometric models. The standard random-effects probit model is applied to the food industry using the balanced panel sample described in Section 3.1. The same modelling procedure is also applied to the balanced panel samples created for businesses belonging to the AgriFF and non-AgriFF subsectors. The model estimates presented here are for overall innovation (i.e. business engaged in any type of innovation). To complement the empirical results and to get a better indication of the effects of the main drivers on overall innovation, the average partial effects (APEs) are included. In all models, the reference firm belonging to the AgriFF subsector (for the

food industry sample only) is small, does not collaborate, has low ICT intensity, has no effective competition, does not use STEM skills, with no debt or equity finance, has export capability and does not have any flexible working arrangements. The reference year is 2007–2008. The pooled probit panel model was also estimated to check the sensitivity of the random-effects probit results. The pooled probit model and multivariate probit model estimates are contained in the Appendix S1, together with the summary statistics for the distributions of the APEs. The empirical results are compared with those of other ABS and DIIS

studies that used cross-sectional modelling and employed data collected in the same survey (i.e. BCS) as well as with the findings coming from previous international food industry studies.

4.1 Findings for the food industry

Table 2 presents the estimated coefficients and their bootstrapped standard errors (SEs) for the standard random-effects probit model of overall innovation propensity. The estimated coefficients for the dummy variables for industry and business size are not significant, contrary to the results of Zouaghi and Sanchez (2016), although the latter covered all Spanish firms in the food sectors. Despite these results, it is important that these two covariates remain in the model because this information is used in the BCS survey design to generate the randomly sampled businesses in the food industry, and the current modelling did not apply any weights. The estimated proportion of the total variance contributed by the panel-level variance component (rho) is significantly different from zero (using the likelihood ratio test) and indicates that this component accounts for more than 44 per cent of the variance of the composite error. These results are consistent with the

What’s driving innovation in small food businesses in Australia? 57

findings of Triguero et al. (2013) and Vancauteren (2016), indicating that controlling for unobserved effects is important in the current analysis. The log likelihood, AIC and BIC results also confirm the goodness of fit of the model compared with the pooled probit model, as exhibited in Table A1 in the Appendix S1. Overall, there is statistical significance at the 5 per cent level for all three

key business characteristics (i.e. collaboration, use of STEM skills and ICT intensity). All play an important role in explaining the innovation behaviour of small businesses in the food industry. All of the APEs are found highly significant except for the APE of moderate ICT intensity which is small compared with the rest, but still significant at the 5 per cent level. The distributions of the estimated APEs are presented in Table C1 in the Appendix S1. We note that all of the distributions are skewed to the left, which implies that there were a significant number of firms where the effects of the drivers were positive, but relatively small in magnitude. Therefore, the

Table 2 Standard random-effects probit regression results for (overall) innovation and average partial effects for selected key drivers of innovation

Variable Coefficient Bootstrap SE APE† SE‡

Innovation (response variable) Subindustry (non-AgriFF) 0.14 0.13 Business size (nonemploying) 0.22 0.14 With collaboration 0.68*** 0.12 0.172*** 0.029

Market competition Minimal 0.50*** 0.18 0.114*** 0.041 Moderate 0.87*** 0.16 0.207*** 0.035 Strong 0.67*** 0.15 0.156*** 0.033

ICT intensity Moderate 0.25** 0.12 0.061** 0.029 High 0.73*** 0.24 0.183*** 0.057 Most intense 0.71*** 0.20 0.177*** 0.048

Used STEM skills 0.58*** 0.13 0.144*** 0.033 Export capability (Local only) �0.33** 0.16 With flexible working arrangements 0.68*** 0.11 0.171*** 0.028 Sought debt and/or equity finance 0.23** 0.11 Financial year 2008–2009 �0.45*** 0.13 2009–2010 �0.50*** 0.12 2010–2011 �0.52*** 0.13

Intercept �1.33*** 0.23 Log likelihood �844.89 AIC 1725.77 BIC 1823.32 Sigma 0.894 0.085 rho 0.444*** 0.047 Number of observations (n) 1668

Note: The asterisks, ***, ** and *, denote significance at the 1%, 5% and 10% levels, respectively. All SEs are given to two significant digits, and the corresponding coefficients are given to the same number of digits behind the decimal points as their SEs. †Average partial effects for selected key drivers of innovation. ‡SEs for APEs are computed using bootstrapping.

58 F. A. Soriano et al.

results suggest that there are potential avenues to intensify the effects of the key drivers on the propensity to innovate among firms. Collaboration is positive and highly significant, implying that food

businesses reporting involvement in any collaborative arrangement are more likely to innovate. This conforms to the findings in ABS (2008), Rotaru (2013), Rotaru et al. (2013), Rotaru and Soriano (2013) and DITR (2006) where collaboration was found to strongly drive overall innovation among Australian businesses. The corresponding APE result shows that after averaging across all small food businesses and time periods, assuming all other variables constant, having some form of collaboration is associated with an increase of more than 17 per cent in the propensity to innovate. The APEs for collaboration at the 25th and 75th percentiles are 15.9 and 19.6 per cent, respectively. This indicates that even after accounting for unobserved heterogeneity, engaging in collaborative arrangements plays an important role for the likelihood of a firm to innovate. This result provides evidence that joint R&D collaboration between small food firms and universities, private and public organisations and other firms plays a key role on business innovation (Grunert et al. 1997; Maietta 2015; Caiazza and Stanton 2016; Ciliberti et al. 2016a,b; De Martino and Magnotti 2017; Sauer 2017; Wixe et al. 2017). This result also implies that small businesses in the Australian food industry can gain external knowledge, acquire technology and improve their technical capability to boost their innovation performance through collaboration. Access to external knowledge through collaborations is an important factor that drives the survival and growth of local food producers (Wixe et al. 2017). Businesses require resources with a broad range of technical and nontechnical skills and capabilities working together to foster innovation (Cunningham et al. 2016). Small food firms usually lack skills and resources; hence, they need external partners/networks to realise innovations (Caiazza and Stanton 2016; Galati et al. 2016). Even horizontal collaboration between firms in the Australian food industry should be encouraged (McAdam et al. 2014). The above results also support the Nelson and Winter (1982) model and the systems of innovation approaches by Nelson (1993) and Lundvall (2010). It is interesting that the results are supportive of the OI models and with the results in Bigliardi and Galati (2013). The use of STEM skills is positively associated with the likelihood of

innovation. Small food businesses that use STEM skills in their production are significantly more likely to engage in any one of the four broad types of innovation than other businesses. These results are consistent with Soriano and Abello (2015), for all business sizes in all industries. Its impact is around 14 per cent with the 25th and 75th percentiles at 13.1 and 16.6 per cent, respectively. The results are consistent with Palangkaraya et al. (2016) on science and research core skills, having significant positive association with any types of innovation in Australian businesses. The significant relationship between employee skills and innovation has also been observed in Huiban and Bouhsina (1998), Avermaete et al. (2004), Capitanio et al. (2009), Smit

What’s driving innovation in small food businesses in Australia? 59

et al. (2015), Vancauteren (2016) and Brown and Roper (2017). This implies that the quality of the human capital in the small food businesses has strong influence on the innovation propensity. A recent Australian farm study exemplified a very strong link between skills and training, innovation and productivity (Xayavong et al. 2015). There is a strong relationship between ICT intensity and the likelihood that

a small food business will undertake innovative activity. The empirical results indicate that, all other things being held constant, moving from moderate to high and more intense ICT increases the likelihood of innovation. They imply that small businesses in the food industry that engage in more sophisticated forms of ICT are significantly more likely to innovate. Averaging across all small food businesses and time periods, assuming all other variables constant, small food businesses using high ICT intensity are 18.3 per cent more likely to innovate than businesses using low ICT intensity. The APEs for high ICT intensity at the 25th and 75th percentiles are 16.6 and 21.0 per cent, respectively. Todhunter and Abello (2011), Rotaru et al. (2013), Tiy et al. (2013) and Soriano and Abello (2015) found similar results in their cross- sectional analyses. The results support the findings of Galati et al. (2016) that ICT plays a crucial role for effective OI performance for food firms as well as the work of Domenech et al. (2014) on the adoption of ICT innovations in the agri-food sector. ICTs have been transforming economic activities in all sectors, particularly farming (including the food industry) for technological innovation and improvement in agricultural output (World Bank 2011; Lamb 2013; Salim et al. 2015). Having flexible working arrangements was found to be significant and had

a strongly positive effect on a firm’s propensity to innovate. This means that a small food business is more likely to innovate if it offers flexible working arrangements for employees. Rotaru (2013) and Palangkaraya et al. (2016) obtained similar results for Australian businesses. The findings also conform with Storey et al. (2002), Michie and Sheehan (2003), Kleinknecht et al. (2006), Martinez-Sanchez et al. (2008), Chung (2009), Zhou et al. (2011) and Kleinknecht et al. (2014). Market competition had a positive and significant association with the

propensity to innovate, indicating that the more competitors that small businesses face, the more likely they are to innovate. This supports the anti- Schumpeterian frameworks. The strongest effect was felt when small food businesses have three to four competitors in the market, which is a similar finding to that of Soames et al. (2011). This highest impact is reflected in the estimated ATE, which is an increase of more than 20 per cent in the probability for small businesses to innovate. The APEs at the 25th and 75th percentiles were also higher at 17.2 and 24.4 per cent, respectively. This complements the result of Smit et al. (2015) where weak association was found between competition and innovation but aligns with the theoretical works of Aghion et al. (2005) and Correa (2012). The following results were obtained for other business characteristics:

60 F. A. Soriano et al.

• The geographic market for the food industry from which the small businesses sell their goods or services played an important role in propensity to innovate. Compared with a firm that operates only locally, expanding the business operation of small firms to overseas markets positively and significantly affected the likelihood to perform any form of innovation. In fact, the recent work of Tuhin (2016) indicated that innovation and export behaviour of Australian small and medium businesses were interrelated and that exporters were 7 per cent to 10 per cent more likely to be innovators, which shows some consistency with the findings of Domenech et al. (2014) and Zouaghi and Sanchez (2016).

• Access to finance is of crucial importance for investing in innovations (Sauer 2017). In this study, small businesses that sought debt and/or equity finance were found significantly more likely to innovate. The lack of financial resources suffered by small agri-food firms influenced innovation, but public funding can boost their innovation capacity (Caiazza and Stanton 2016; and De Martino and Magnotti 2017). This result shows the need of small food businesses in Australia for continued government assistance to help sustain their innovation activities.

Inclusion of year effects in a short balanced panel is important because it controls for the unexpected variation or special events that may affect the response variable. In terms of the year effects, the results indicate that, all other things being held constant, the association is negative and significant moving onwards from 2007 to 2008. This perhaps reflects the effect on a firm’s likelihood to engage in any form of innovation of the global financial crisis that occurred in 2008. A similar negative impact was also evident from the work of Zouaghi and Sanchez (2016). Table B1 in the Appendix S1 presents the results for the multivariate probit

modelling. The estimated coefficients for collaboration, the use of resources with STEM skills, ICT, flexible working arrangements, competitiveness and business size for each of the types of innovation are in line with our expectations (i.e. positive and significant). Export capability was found significant in marketing method innovation, whereas access to finance through debt and equity was only significant for operational process innovation. Another key result of the multivariate probit model is in the lower part of the same table where correlations between the residuals in the four equations (Eqn 7) are reported. All of the estimated correlations are significantly positive, which indicates that small food industry businesses combine different types of innovation. The significance might be associated with possible omitted variables (e.g. R&D intensity; firm age).

4.2 Findings for food industry subsectors

In this subsection, we expand the food industry model by dividing the food industry balanced sample into two subsamples, the AgriFF panel (948

What’s driving innovation in small food businesses in Australia? 61

T a b le

3 S ta n d a rd

ra n d o m -e ff ec ts

p ro b it re g re ss io n re su lt s fo r (o v er a ll ) in n o v a ti o n a n d a v er a g e p a rt ia l eff

ec ts fo r se le ct ed

k ey

b u si n es s ch a ra ct er is ti cs ,

b y fo o d in d u st ry

su b se ct o r

V a ri a b le s

A g ri F F

N o n -A

g ri F F

C o effi

ci en t

B o o ts tr a p S E

A P E †

S E ‡

C o effi

ci en t

B o o ts tr a p S E

A P E †

S E ‡

In n o v a ti o n (r es p o n se

v a ri a b le )

B u si n es s si ze

(n o n em

p lo y in g )

0 .0 9

0 .2 2

0 .3 2

0 .2 2

W it h co ll a b o ra ti o n

0 .7 1 * * *

0 .1 6

0 .1 7 9 * * *

0 .0 4 4

0 .7 0 * * *

0 .2 3

0 .1 7 7 * * *

0 .0 5 3

M a rk et

co m p et it io n

M in im

a l

0 .3 0 *

0 .1 8

0 .0 6 3

0 .0 4 5

0 .7 1 * *

0 .3 5

0 .1 7 *

0 .1 0

M o d er a te

0 .8 9 * * *

0 .1 4

0 .2 1 1 * * *

0 .0 4 6

0 .8 5 * * *

0 .3 0

0 .2 1 0 * * *

0 .0 8 3

S tr o n g

0 .7 2 * * *

0 .1 7

0 .1 6 5 * * *

0 .0 3 8

0 .6 1 * *

0 .3 1

0 .1 5 0 *

0 .0 8 6

IC T in te n si ty

(h ig h to

m o st

in te n se )

0 .5 0 * *

0 .2 4

0 .1 2 3 * *

0 .0 5 2

0 .6 2 * * *

0 .1 6

0 .1 6 1 * * *

0 .0 5 0

U se d S T E M

sk il ls

0 .6 3 * * *

0 .1 5

0 .1 5 7 * * *

0 .0 4 0

0 .5 1 * * *

0 .1 5

0 .1 2 7 * * *

0 .0 4 6

E x p o rt

ca p a b il it y (L o ca l o n ly )

�0 .3 0

0 .2 0

�0 .3 2

0 .2 0

W it h fl ex ib le

w o rk in g a rr a n g em

en ts

0 .7 1 * * *

0 .2 0

0 .1 7 4 * * *

0 .0 3 7

0 .6 3 * * *

0 .1 1

0 .1 6 3 * * *

0 .0 3 9

S o u g h t d eb t a n d /o r eq u it y fi n a n ce

0 .1 9

0 .2 0

0 .2 7

0 .2 1

F in a n ci a l y ea r

2 0 0 8 – 2 0 0 9

�0 .3 3 *

0 .1 8

�0 .5 2 * * *

0 .1 5

2 0 0 9 – 2 0 1 0

�0 .4 4 * *

0 .2 2

�0 .5 2 * * *

0 .1 8

2 0 1 0 – 2 0 1 1

�0 .4 0 * * *

0 .1 4

�0 .5 8 * *

0 .2 3

In te rc ep t

�1 .2 2 * * *

0 .3 3

�1 .0 2 * * *

0 .2 8

L o g li k el ih o o d

�4 6 5 .1 8

�3 7 8 .1 4

A IC

9 4 8 .3 5

7 7 4 .2 7

B IC

9 9 2 .0 4

8 1 5 .4 9

S ig m a

0 .8 5 3

0 .0 5 8

0 .9 4

0 .1 6

R h o

0 .4 2 1 * * *

0 .0 3 3

0 .4 6 9 * * *

0 .0 8 5

N u m b er

o f o b se rv a ti o n s (n )

9 4 8

7 2 0

N o te : T h e a st er is k s,

* * * , * * a n d * , d en o te

si g n ifi ca n ce

a t th e 1 %

, 5 %

a n d 1 0 %

le v el s,

re sp ec ti v el y . A ll S E s a re

g iv en

to tw

o si g n ifi ca n t d ig it s,

a n d th e co rr es p o n d in g

co effi

ci en ts a re

g iv en

to th e sa m e n u m b er

o f d ig it s b eh in d th e d ec im

a l p o in ts a s th ei r S E s. † A v er a g e p a rt ia l eff

ec ts fo r se le ct ed

k ey

d ri v er s o f in n o v a ti o n . ‡ S E s fo r A P E s a re

co m p u te d u si n g b o o ts tr a p p in g .

62 F. A. Soriano et al.

observations) and the non-Agriff panel (720 observations). In the subsector models, the categories in the ICT intensity variable are combined to form two binary dummy variables because of the presence of high correlations between the original four categories in the subsector samples. This is brought about by the reduced number of observations in the subsector samples. The results using the standard random-effects probit modelling are presented in Table 3. The results of the random-effects models are consistent with our expec-

tations. The direction and sign of the estimated coefficients for both dummy variables for AgriFF and non-AgriFF are consistent with the food industry results. However, the coefficients of a few explanatory variables are not statistically significant, as for the food industry results. Coefficients of the export capability and finance variables are not significant in both samples. In addition, the estimated coefficients in the AgriFF for minimal market competition and the year 2008–2009 are significant at the 10 per cent level. The coefficient of the business size variable is not significant at the 10 per cent level, as in the case for the food industry. The estimated proportions of the total variance contributed by the panel-

level variance component (rho) are found significantly different from zero (using the likelihood ratio test). The models for AgriFF and for non-AgriFF account for more than 42 per cent of the variance of the composite error, the percentage for the latter subsector being somewhat larger (47 per cent).These results indicate that accounting for unobserved effects is important in the subsector analyses. The log likelihood, AIC and BIC results also confirm the goodness of fit of the subsector models compared with the pooled probit models, as exhibited in Table A2 in the Appendix S1. The five key factors (collaboration, ICT intensity, flexible working

arrangements, facing market competition and use of STEM skills) again were highly significant in contributing to increase innovation among small food businesses in both subsectors. Year effects, in general, remained negative and statistically significant. The global financial crisis had greater effect on the likelihood to innovate for small food businesses in the non-AgriFF model than for small food businesses in the AgriFF model. The overall results of the multivariate probit modelling for the two

subsectors are consistent with the above results. Correlations between the disturbances for all the innovation outcomes were also positive and significant implying that small food firms even at the subsector level do combine different types of innovation when innovating. All APEs in the AgriFF model were found to be statistically significant

except for the case of minimal market competition, which is smaller in magnitude than the rest. Two of the APEs (minimal and strong market competition) in the non-AgriFF model are significant at the 5 per cent level, whereas the rest are found highly statistically significant (at the 1 per cent level). After averaging across all small food businesses in each subsector and time periods and assuming all other variables constant, the degree of

What’s driving innovation in small food businesses in Australia? 63

association between having intense ICT and innovation is stronger in the non-AgriFF subsector, whereas the effect on innovation propensity is stronger in the AgriFF subsector when small food businesses use STEM skills and provide flexible working arrangements for employees. The impact of collaboration is similar in both subsectors which is approximately 18 per cent. The distributions of the APEs for the subsector modelling are presented in Tables C2 and C3 in the Appendix S1.

5. Conclusions

This study investigates the key drivers of innovation among small businesses in the Australian food industry, which is an important part of the country’s regional economies. While the modelling technique used to identify the factors that play important roles in influencing the innovation behaviour of small food businesses and their strength of association is well established in the literature (see Wooldridge 2005, 2010, 2013 and Greene 2012), estimating robust standard errors for impact summary measures called ‘average partial effects’ simulation using modern bootstrapping procedures is a novelty in this work. The empirical work utilises firm-level information collected through the ABS BCS (2006–2007 to 2010–2011 ABS BLD CURF) and accessed via the ABS Data Laboratory. In this paper, we observed the connection between collaboration, the use of

STEM skills, the use of ICT, having flexible working arrangements, degree of market competition and innovation among small businesses in the Australian food industry. The analyses focused on one aspect of innovation – the propensity of the businesses to innovate. Other possible drivers of innovation, such as business size, export capability and finance sought, were also examined. Overall, we found that innovation is an essential element in small food businesses primary activities; also in Avermaete et al. (2003), Baregheh et al. (2012), Caiazza and Stanton (2016), De Martino and Magnotti (2017), and Wixe et al. (2017), and the results around associations between the above drivers and innovation propensity were similar with most of the ABS cross- sectional studies on innovation. One of the highlights in the analyses was the significant and positive

association between collaboration and business innovation among small food businesses, where we found that businesses engaging in any form of collaborative arrangements were more likely to innovate. This is similar to the results of most of the previous international studies. The adoption of OI practices in the small food industry in Australia should be supported and encouraged by the DIIS. The work of Huiban and Bouhsina (1998), Avermaete et al. (2004), Capitanio et al. (2009), Smit et al. (2015), Van- cauteren (2016), and Brown and Roper (2017) has shown that business innovation is influenced by the workforce’s skills and the quality of human capital in the food industry. In this paper, our empirical results also indicate that small food businesses using STEM skills are significantly more likely to

64 F. A. Soriano et al.

innovate. The above findings are currently backing the mission of the Food and Agribusiness Growth Centre – the FIAL – to build capability and encourage collaboration and innovation in the Australian food sector (FIAL 2017). The findings in our paper also raise important policy implications for

growing the food industry in Australia. It is supporting the Industry Innovation and Competitiveness Agenda for small businesses to grow and become competitive both nationally and internationally. The Australian Government appears on the right track in its commitment to create favourable conditions and an environment in small food businesses to utilise the emerging ICT technologies in their business activities; offer flexible working arrangements to their employees; support access to new interna- tional markets and increase industry competitiveness. Moreover, the unique sample data for the Australian food industry have

enabled us to undertake separate analyses for businesses belonging to the AgriFF and the non-AgriFF subsectors of the food industry. Additional findings, obtained in these subsector analyses, indicate that the degree of association between having intense ICT and innovation is stronger in the non-AgriFF subsector. On the other hand, the impact on innovation is stronger in the AgriFF subsector when small businesses use STEM skills and provide flexible working arrangements to employees. The impact of collab- oration is positive and statistically significant and is similar in magnitude in both subsectors. Further, OI has shown its importance at the subindustry level. The main output-based measure of innovation used in this paper was

overall innovation. However, we also found positive and significant corre- lation structures between the four types of innovation: new goods and services, new organisational processes, new operational processes and new marketing methods using multivariate probit. Future work could expand the analyses in the small food businesses by considering each type of innovation separately and examining how sensitive the impacts of the drivers are to the different types of innovation, as has been done in the recent work of Smit et al. (2015), Zouaghi and Sanchez (2016) and Wixe et al. (2017). In addition, there is strong evidence suggesting that persistence of innovation is an important characteristic of successful firms (Rotaru 2013; Rotaru and Soriano 2013; Triguero et al. 2013; Tavassoli and Karlsson 2015), and this suggests that the causes of persistence in innovation in the small food businesses are worthy of further investigation. From the modelling perspec- tive of innovation persistence, it is desirable to account for correlation between the firm-specific effects and the regressors, because there may exist time-varying regressors across the periods. Now that we have established what drives small food businesses to

innovate, a next step would be to determine whether these innovating businesses are becoming more productive over time. A recent study by Palangkaraya et al. (2015) showed that Australian SMEs that previously

What’s driving innovation in small food businesses in Australia? 65

introduced innovations had an annual productivity increase of around 2.7 per cent higher than noninnovating SMEs and could further raise their productivity by 4.4 per cent if they were to collaborate. The recent work of Khan et al. (2017) in Australia’s broadacre agriculture revealed strong link between R&D and productivity. Whether this is so for small food businesses in Australia is worth investigating to provide additional evidence to support the government’s policies and investments in growing Australian food businesses.

References

ABS (2001). Australian Standard Classification of Education (ASCED) 2001. Australian

Bureau of Statistics, Canberra, ACT. ABS (2002). Small Business in Australia 2001. Australian Bureau of Statistics, Canberra, ACT. ABS (2006). Australian and New Zealand Standard Industrial Classification 2006. Australian Bureau of Statistics, Canberra, ACT.

ABS (2008). Key Aspects of Innovation in Australian Businesses: Micro-Data Analysis of the 2003 and 2005 Innovation Surveys. ABS Submission to the Review of the National Innovation System. Australian Bureau of Statistics, Canberra, ACT.

ABS (2013). Microdata: Business Longitudinal Database, Australia 2006–07 to 2010–11. Australian Bureau of Statistics, Canberra, ACT.

ABS (2015a). Selected Characteristics of Australian Business 2013–14. Australian Bureau of Statistics, Canberra, ACT.

ABS (2015b). Australian Industry 2013–14. Australian Bureau of Statistics, Canberra, ACT. ABS (2015c). Summary of IT Use and Innovation in Australian Business 2013–14. Australian Bureau of Statistics, Canberra, ACT.

ABS (2015d). Summary of IT Use and Innovation in Selected Growth Sectors Australia 2013–14. Australian Bureau of Statistics, Canberra, ACT.

ABS (2015e). Information Paper: Construction of the Expanded Analytical Business Longitu-

dinal Database 2001–02 to 2012–13. Australian Bureau of Statistics, Canberra, ACT. ABS (2017). Selected Characteristics of Australian Business 2015–16. Australian Bureau of Statistics, Canberra, ACT.

ABS (2018). Counts of Australian Businesses, including Entries and Exits, June 2013 to June 2017. Australian Bureau of Statistics, Canberra, ACT.

Aghion, P. and Howitt, P. (2009). The Economics of Growth. MIT Press, Cambridge, MA.

Aghion, P., Bloom, N., Blundell, R., Griffith, R. and Howitt, P. (2005). Competition and innovation: an inverted-U relationship, Quarterly Journal of Economics 120, 701–728.

Aghion, P., Bergeaud, A., Lequin, M. and Melitz, M. (2018). The impact of exports on innovation: theory and evidence, NBER Working Paper No. 24600. National Bureau of

Economic Research, Cambridge, MA. Atkinson, J. (1984). Manpower strategies for flexible organisations, IMS Report No. 89. Institute of Manpower Studies, University of Sussex, Brighton.

Australian Government (2015). Australian innovation system report 2015. Office of the Chief Economist, Department of Industry, Innovation and Science, Canberra, ACT.

Australian Government (2017). National Innovation and Science Agenda. Department of

Industry, Innovation and Science, Canberra, ACT. Avermaete, T., Viaene, J., Morgan, E.J. and Crawford, N. (2003). Determinants of innovation in small food firms, European Journal of Innovation Management 6, 8–17.

Avermaete, T., Viaene, J., Morgan, E.J., Pitts, E., Crawford, N. and Mahon, D. (2004). Determinants of product and process innovation in small food manufacturing firms, Trends in Food Science and Technology 15, 474–483.

66 F. A. Soriano et al.

AWPA (2013). Human Capital and Productivity: Literature Review. Australian Workforce and Productivity Agency, Canberra, ACT.

Baregheh, A., Rowley, J., Sambrook, S. and Davies, D. (2012). Innovation in food sector SMEs, Journal of Small Business and Enterprise Development 19, 300–321.

Beatson, M. (1995). Labour market flexibility, Research Series No. 48. Employment Department, Moorfoot, Sheffield.

Becker, G.S. (1994). Human Capital: A Theoretical and Empirical Analysis with Special

Reference to Education, 3rd edn. The University of Chicago Press, National Bureau of Economic Research, Cambridge, MA.

Bhattacharya, M. and Bloch, H. (2004). Determinants of innovation, Small Business

Economics 22, 155–162. Bigliardi, B. and Galati, F. (2013). Models of adoption of open innovation within the food industry, Trends in Food Science & Technology 30, 16–26.

Bigliardi, B. and Galati, F. (2016). Which factors hinder the adoption of open innovation in SMEs?, Technology Analysis & Strategic Management 28, 869–885.

Brown, P. and Roper, S. (2017). Innovation and networks in New Zealand farming, Australian Journal of Agricultural and Resource Economics 61, 422–442.

Caiazza, R. and Stanton, J. (2016). The effect of strategic partnership on innovation: an empirical analysis, Trends in Food Science & Technology 54, 208–212.

Capitanio, F., Coppola, A. and Pascucci, S. (2009). Indications for drivers of innovation in the

food sector, British Food Journal 111, 820–838. Capitanio, F., Coppola, A. and Pascucci, S. (2010). Product and process innovation in the Italian food industry, Agribusiness 26, 503–518.

Cappellari, L. and Jenkins, P. (2003). Multivariate probit regression using simulated maximum likelihood, The Stata Journal 3, 278–294.

Chamberlain, G. (1984). Panel data, in Griliches, Z. and Intriligator, M. (eds), Handbook of Econometrics 2. North-Holland, Amsterdam, pp. 1,247–1,318.

Chaustowski, R. and Dolman, S. (2016). Australia’s food and agribusiness sector – data profile, Economic and Analytical Services Division Paper. Department of Industry, Innovation and Science, Canberra, ACT.

Chesbrough, H.W. (2003). The era of open innovation, MIT Sloan Management Review 44, 35–41. Massachusetts Institute of Technology, Cambridge, MA.

Chung, H. (2009). Flexibility for Whom? Working Time Flexibility Practices of European

Companies. Tilburg University, Tilburg. Ciliberti, S., Carraresi, L. and Br€oring, S. (2016a). Drivers of innovation in Italy: food versus pharmaceutical industry, British Food Journal 118, 1,292–1,316.

Ciliberti, S., Carraresi, L. and Br€oring, S. (2016b). External knowledge sources as drivers for cross-industry innovation in the Italian food sector: does company size matter?, Interna- tional Food and Agribusiness Management Review 9, 77–98.

Correa, J.A. (2012). Innovation and competition: an unstable relationship, Journal of Applied

Econometrics 27, 160–166. Cunningham, S., Theilacker, M., Gahan, P., Callan, V. and Rainnie, A. (2016). Skills and capabilities for Australian enterprise innovation, Report for the Australian Council of

Learned Academies. Melbourne, VIC. De Martino, M. and Magnotti, F. (2017). The innovation capacity of small food firms in Italy, European Journal of Innovation Management 3, 362–383.

DIIS (2008). Collaborating to a purpose, Review of the Cooperative Research Centres Program. Department of Industry, Innovation and Science, Canberra, ACT.

DIIS (2017). Food and Agribusiness Industry. Department of Industry, Innovation and Science, Canberra, ACT.

DIIS (2018a). Australia 2030: Prosperity Through Innovation. Department of Industry, Innovation and Science, Canberra, ACT.

What’s driving innovation in small food businesses in Australia? 67

DIIS (2018b). Industry-research Collaboration. Department of Industry, Innovation and Science, Canberra, ACT.

DIISR (2009). Powering Ideas: An Innovation Agenda for the 21st Century. Department of Industry, Innovation, Science and Research, Canberra, ACT.

DITR (2006). Collaboration and Other Factors Influencing Innovation Novelty in Australian Businesses: An Econometric Analysis. Department of Industry, Tourism and Resources, Canberra, ACT.

Domenech, J., Martinez-Gomez, V. and Mas-Verdu, F. (2014). Location and adoption of ICT innovations in the agri-food industry, Applied Economics Letters 21, 421–424.

DPMC (2014). Industry Innovation and Competitiveness Agenda: An Action Plan for a Stronger

Australia. Department of the Prime Minister and Cabinet, Canberra, ACT. FIAL (2017). Food and Agribusiness Sector Competitiveness Plan. Food Innovation Australia Ltd, Melbourne, VIC.

Gago, D. and Rubalcaba, L. (2007). Innovation and ICT in service firms: towards a multidimensional approach for impact assessment, Journal of Evolutionary Economics 17, 25–44.

Galati, F., Bigliardi, B. and Petroni, A. (2016). Open innovation in food firms: implementation

strategies, drivers and enabling factors, International Journal of Innovation Management 20, 1650042.

Grant, R.M. (1991). A resource based theory of competitive advantage, California Manage-

ment Review 3, 114–135. Greene, W.H. (2012). Econometric Analysis, 7th edn. Prentice Hall, Upper Saddle River, NJ. Gretton, P., Gali, J. and Parham, D. (2003). The Effects of ICTs and Complementary

Innovations on Australian Productivity Growth. Productivity Commission, Canberra, ACT. Griffiths, W. and Webster, E. (2009). What governs firm-level R&D: internal or external factors?, Melbourne Institute Working Paper Series 13. University of Melbourne, Melbourne, VIC.

Grossman, G.M. and Helpman, E. (1991). Quality ladders in the theory of growth, The Review of Economic Studies 58, 43.

Grunert, K., Harmsen, H., Meulenberg, M., Kuiper, E., Ottowitz, T., Declerck, F., Traill, B.

and G€oransson, G. (1997). A framework for analysing innovation in the food sector, in Traill, B. and Grunert, K. (eds), Product and Process Innovation in the Food Industry. Blackie Academic and Professional, London, pp. 1–37.

Hendrickson, L., Bucifal, S., Balaguer, A. and Hansell, D. (2015). The employment dynamics of Australian entrepreneurship, Office of the Chief Economist Research Paper 4/2015. Department of Industry, Innovation and Science, Canberra, ACT.

Hensher, D.A., Rose, J.M. and Greene, W.H. (2015). Applied Choice Analysis, 2nd edn. Cambridge University Press, New York.

Henttonen, K. and Lehtim€aki, H. (2017). Open innovation in SMEs: collaboration modes and strategies for commercialization in technology-intensive companies in forestry industry,

European Journal of Innovation Management 20, 329–347. Huiban, J.-P. and Bouhsina, Z. (1998). Innovation and the quality of labour factor: an empirical investigation in the French food industry, Small Business Economics 10, 389–400.

IP Australia (2014). The Australian Food Industry: A Patent Analytics Report. Intellectual Property Australia, Canberra, ACT.

Khan, F., Salim, R., Bloch, H. and Islam, N. (2017). The public R&D and productivity growth

in Australia’s broadacre agriculture: is there a link?, Australian Journal of Agricultural and Resource Economics 61, 285–303.

Kleinknecht, A., Oostendorp, R.M., Pradhan, M.P. and Naastepad, C.W.M. (2006). Flexible labour, firm performance and the Dutch job creation miracle, International Review of

Applied Economics 20, 171–187. Kleinknecht, A., van Flore, N., Schaik, F.N. and Zhou, H. (2014). Is flexible labour good for innovation? Evidence from firm-level data, Cambridge Journal of Economics 38, 1,207–1,219.

68 F. A. Soriano et al.

K€ollinger, P. (2005). Why IT matters: an empirical study of e-business usage, innovation, and firm performance, Discussion Papers of DIW Berlin 495. German Institute for Economic Research, Berlin.

Lamb, D. (2013). Game changer: the role of broadband connectivity in Australian farms and why

we need to get this right?, Grains Research and Development Corporation Update Paper. Grains Research and Development Corporation, Canberra, ACT.

Lundvall, B.-A. (2010). National Systems of Innovation: Toward a Theory of Innovation and

Interactive Learning. Anthem Press, New York, NY. Maietta, O.W. (2015). Determinants of university-firm R&D collaboration and its impact on innovation: a perspective from a low-tech industry, Research Policy 44, 1,341–1,359.

Marangos, S. and Warren, L. (2017). A mapping for managers: open innovation for R&D intensive SMEs in the life science sector, European Journal of Innovation Management 20, 210–229.

Martinez-Sanchez, A., Vela-Jimenez, M.J., Perez-Perez, M. and de-Luis-Carnicer, P. (2008). Workplace flexibility and innovation: the moderator effect of international cooperation, Personnel Review 37, 647–665.

McAdam, M., McAdam, R., Dunn, A. and McCall, C. (2014). Development of small and

medium-sized enterprise horizontal innovation networks: UK agri-food sector study, International Small Business Journal 32, 830–853.

Michie, J. and Sheehan, M. (2003). Labour market deregulation, ‘flexibility’ and innovation,

Cambridge Journal of Economics 27, 123–143. Michie, J. and Sheehan-Quinn, M. (2001). Labour market flexibility, human resource management and corporate performance, British Journal of Management 12, 287–306.

Mundlak, Y. (1978). On the pooling of time series and cross section data, Econometrica 46, 69– 85.

Nelson, R.R. (1993). National Innovation Systems: A Comparative Analysis. Oxford University Press, Oxford, New York.

Nelson, R.R. and Winter, S.G. (1982). An Evolutionary Theory of Economic Change. Harvard University Press, Cambridge, MA.

OECD (2003). A proposed classification of ICT goods, Paper prepared by the Working Party on

Indicators for the Information Society. Organisation for Economic Co-operation and Development, Paris.

OECD (2005a). OECD SME and Entrepreneurship Outlook 2005. Organisation for Economic

Co-operation and Development, Paris. OECD (2005b). The measurement of scientific and technological activities: Guidelines for collecting and interpreting innovation data, Oslo manual, Paper prepared by the Working

Party of the National Experts on Scientific and Technology Indicators. Organisation for Economic Co-operation and Development, Paris.

Palangkaraya, A., Spurling, T. and Webster, E. (2015). Does innovation make (SME) firms more productive?, in Moore, A. and Simon, J. (eds), Small Business Conditions and Finance.

Reserve Bank of Australia, Sydney, NSW, pp. 181–199. Palangkaraya, A., Spurling, T. and Webster, E. (2016). What Drives Firm Innovation? A Review of the Economics Literature. Centre for Transformative Innovation, Swinburne University of

Technology, Melbourne, VIC. Parker, T. (2017). The DataLab of the Australian bureau of statistics, The Australian Economic Review 50, 478–483.

Reilly, P.A. (2001). Flexibility at Work: Balancing the Interest of Employer and Employee. Gower Publishing Limited, Hampshire.

Rogers, M. (2004). Networks, firm size and innovation, Small Business Economics 22, 141–153. Rotaru, C.I. (2013). Discrete choice panel data modelling using the ABS business longitudinal

database. ABS Methodology Research Papers. Australian Bureau of Statistics, Canberra, ACT.

What’s driving innovation in small food businesses in Australia? 69

Rotaru, C. and Soriano, F. (2013). Flexible working arrangements, collaboration, ICT and innovation: A panel data analysis, Paper presented at the 2013 Economic Measurement Group (EMG) Workshop. UNSW, Sydney, NSW.

Rotaru, C., Dzhumasheva, S. and Soriano, F. (2013). Propensity score matching: an

application using the ABS business characteristics survey. ABS Methodology Research Papers. Australian Bureau of Statistics, Canberra, ACT.

Salim, R., Khan Mamun, S.A. and Hassan, K. (2015). Role of communication technologies in

broadacre agriculture in Australia: an empirical analysis using panel data, Australian Journal of Agricultural and Resource Economics 60, 243–264.

Sauer, J. (2017). Estimating the link between farm productivity and innovation in the

Netherlands, OECD Food, Agriculture and Fisheries Papers 102. OECD Publishing, Paris. Schmookler, J. (1966). Invention and Economic Growth. Harvard University Press, Cambridge, MA.

Schumpeter, J. (1942). Capitalism, Socialism, and Democracy. Harper and Brothers Publishers, New York.

Smit, M.J., Abreu, M.A. and de Groot, H.L.F. (2015). Micro-evidence on the determinants of innovation in the Netherlands: the relative importance of absorptive capacity and

agglomeration externalities, Papers in Regional Science 94, 249–272. Smith, R. and Hendrickson, L. (2016). The effect of age on Australian small-to-medium enterprises, Office of the Chief Economist Research Paper 1/2016. Department of Industry,

Innovation and Science, Canberra, ACT. Soames, L., Brunker, D. and Talgaswatta, T. (2011). Competition, innovation and produc- tivity in Australian businesses. ABS Methodology Research Papers. Australian Bureau of

Statistics, Canberra, ACT. Soriano, F. and Abello, R. (2015). Modelling the relationships between the use of STEM skills, collaboration, R&D, and innovation among Australian businesses, Australian Journal of Labour Economics 18, 345–374.

Srholec, M. (2016). Persistence of cooperation on innovation: econometric evidence from panel micro data, Prague Economic Papers 25, 53–70.

Storey, J., Quitas, P., Taylor, P. and Fowle, W. (2002). Flexible employment contracts and

their implications for product and process innovation, International Journal of Human Resource Management 13, 1–18.

Tavassoli, S. and Karlsson, C. (2015). Persistence of various types of innovation analysed and

explained, Research Policy 44, 1,887–1,901. Tirole, J. (1988). The Theory of Industrial Organisation. MIT Press, Cambridge, MA. Tiy, L., Berry, O. and Taylor, D. (2013). Business innovation and the use of information and

communications technology – an update. ABS Methodology Research Papers. Australian Bureau of Statistics, Canberra, ACT.

Todhunter, J. and Abello, R. (2011). Business innovation and the use of information and communications technology. ABS Methodology Research Papers. Australian Bureau of

Statistics, Canberra, ACT. Toner, P. (2011). Workforce skills and innovation: an overview of major themes in the literature, STI Working Paper Series. Directorate for Science, Technology and Industry, OECD, Paris.

Triguero, A., Corcoles, D. and Cuerva, M.C. (2013). Differences in innovation between food and manufacturing firms: an analysis of persistence, Agribusiness 29, 273–292.

Tuhin, R. (2016). Modelling the relationship between innovation and exporting: evidence from

Australian SMEs, Office of the Chief Economist Research Paper 3/2016. Department of Industry, Innovation and Science, Canberra, ACT.

Tuhin, R. and Swanepoel, J. (2017). Export behaviour and business performance: evidence from Australian microdata, Office of the Chief Economist Research Paper 7/2016. Department of

Industry, Innovation and Science, Canberra, ACT. Vancauteren, M. (2016). The effects of human capital, R&D and firm’s innovation on patents: a panel study on Dutch food firms, The Journal of Technology Transfer, 43, 901–922.

70 F. A. Soriano et al.

Wixe, S., Nilsson, P., Naldi, L. and Westlund, H. (2017). Disentangling innovation in small food firms: The role of external knowledge, support, and collaboration, The Royal Institute of Technology, Centre of Excellence for Science and Innovation Studies (CESIS) Electronic Working Paper Series 44. Sweden.

Wong, M., Page, D., Abello, R. and Pang, K.P. (2007). Explorations of innovation and business performances using linked firm-level data. ABS Methodology Research Papers. Australian Bureau of Statistics, Canberra, ACT.

Wooldridge, J.M. (2005). Simple solutions to the initial conditions problem in dynamic, nonlinear panel data models with unobserved heterogeneity, Journal of Applied Economet- rics 20, 39–54.

Wooldridge, J.M. (2010). Econometric Analysis of Cross Section and Panel Data. MIT Press, Cambridge, MA.

Wooldridge, J.M. (2013). Correlated random effects panel data models, IZA European Summer

School in Labor Economics Course Material. Buch/Ammersee, Germany. World Bank (2011). ICT in agriculture: connecting smallholders to knowledge, networks, and institutions, The World Bank Report No. 64605. Washington, DC.

Xayavong, V., Kingwell, R. and Islam, N. (2015). How training and innovation link to farm

performance: a structural equation analysis, Australian Journal of Agricultural and Resource Economics 60, 227–242.

Zhou, H., Dekker, R. and Kleinknecht, A. (2011). Flexible labour and innovation

performance: evidence from longitudinal firm-level data, Industrial and Corporate Change 20, 941–968.

Zouaghi, F. and Sanchez, M. (2016). Has the global financial crisis had different effects on

innovation performance in the agri-food sector by comparison to the rest of the economy?, Trends in Food Science & Technology 50, 230–242.

Supporting Information

Additional supporting information may be found online in the Supporting Information section at the end of the article Appendix S1. (A) Pooled probit model results (sensitivity analysis), (B)

Multivariate probit model results, (C) Selected summary statistics for the distribution of the APE estimates (random effects probit model).

What’s driving innovation in small food businesses in Australia? 71

Copyright of Australian Journal of Agricultural & Resource Economics is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.