1 GROWTH, EXPORTING AND INNOVATION IN MANUFACTURING SMEs: EVIDENCE FROM AUSTRALIA’S BUSINESS LONGITUDINAL SURVEY Professor Richard G.P. McMahon, Head, School of Commerce, The Flinders University of South Australia, GPO Box 2100, Adelaide South Australia 5001. Telephone: +61 8 82012840 Facsimile: +61 8 82012644 Email: Richard.McMahon@flinders.edu.au SCHOOL OF COMMERCE RESEARCH PAPER SERIES: 00-10 ISSN: 1441-3906 Acknowledgments Permission from the Australian Statistician to use confidentialised data from the federal government’s Business Longitudinal Survey is gratefully acknowledged. Responsibility for the findings of this research lies solely with the author. 2 GROWTH, EXPORTING AND INNOVATION IN MANUFACTURING SMEs: EVIDENCE FROM AUSTRALIA’S BUSINESS LONGITUDINAL SURVEY Abstract Two broad hypotheses are examined in this paper. The first is that, in small and medium-sized enterprises (SMEs), linkages exist between growth, exporting and innovation that are statistically significant and persistent over time. Overall support is found for this hypothesis. The second hypothesis examined is that knowledge of exporting and innovation behaviour can be reliably used to anticipate SME growth potential for research and policy purposes. Support for this hypothesis is less compelling. 3 INTRODUCTION A broad definition of a growth SME (small or medium-sized enterprise) suggested by McMahon et al. (1993, p. 15) is as follows: In general terms, a smaller growth enterprise is considered to be an owner-managed concern experiencing on-going, significant and often rapid increases in some or all of the usual indicators of size such as sales revenues, assets, profits (possibly) and number of employees; and which may also be moving towards greater product, geographical or technological diversity. This definition does not equate growth with just increases in enterprise size, however measured. Greater diversity of operations – for example, arising from internationalisation and/or innovation – is seen as a concomitant with growth. Reference to greater geographical diversity posits a linkage between SME growth and export development as a growth strategy. Reference to greater technological diversity posits a further linkage between SME growth and technological innovation as a growth strategy. A report by McKinsey & Company (1994, pp. vii-viii) to the Australian Manufacturing Council reinforces the impression of important linkages between manufacturing SME growth, exporting and innovation: Our aim . . . has been to understand the nature of the relationships or linkages that are important for sustaining the growth of our manufacturing exporters – relationships that drive successful innovation and performance improvement. . . . exporting and innovativeness are closely intertwined. We do not know in which direction the causation goes but innovative firms are more likely to export, and exporting provides opportunities, especially in gaining access to ideas, which are essential for innovativeness. Government policy that aims to enhance both exporting and innovativeness is mutually reinforcing. Push on one and the economy also gets the advantage of the other. Encouraging firms to export might well be the single best measure to stimulate innovation in Australian manufacturing. 4 The recent availability of data from Australia’s Business Longitudinal Survey (BLS) provides a promising new opportunity to explore the linkages so identified. This paper builds upon a study previously undertaken by the author (McMahon, forthcoming) as part of an on-going research effort to derive, characterise and employ an empirically-based development taxonomy for SMEs operating as proprietary companies in the manufacturing sector, using panel data now available from the BLS. The principal objective in the paper is to examine the existence, persistence and usefulness of conjectured linkages between SME growth, exporting and innovation amongst Australian manufacturers. The paper proceeds as follows. After briefly outlining key findings of the author’s previous study, and of some prior research on growth, exporting and innovation amongst Australian SMEs, the current research method is outlined. Thereafter, findings of the research are presented, followed by conclusions arising from this investigation. PRIOR RESEARCH SME Growth In the author’s previous research (McMahon, forthcoming), exploratory cluster analysis was used with key enterprise age, size and growth variables to discover if there appear to be any stable development pathways evident in the BLS panel data. Each of four annual data collections for the on-going longitudinal panel of 871 manufacturing SMEs was separately examined using cluster analysis. Comparisons were then made of cluster analysis outcomes over time. Using the clusters as markers or signposts, three relatively stable SME development pathways were discernible in the longitudinal panel results – low, moderate and high growth. The low growth development pathway appears to account for approximately 70 per cent of SMEs in the panel. The moderate growth pathway seems to be followed by roughly 25 per cent of the panel. And around 5 per cent of the panel look to lie on the high growth pathway, which is in accord with the observed rarity of substantial growth amongst 5 SMEs world-wide (McMahon et al., 1993). Differences between the identified SME development pathways in terms of enterprise age, size and growth variables are highly significant in a statistical sense, thus underpinning confidence in the development taxonomy. It would appear that the development pathways and the pace of SME development (over 20 years or so) in the McMahon (forthcoming) study match reasonably well with those in earlier research of a similar nature undertaken by Hanks et al. (1993). Both development models seem to lead towards the same range of SME configurations that are widely recognised in the relevant research literature (McMahon et al., 1993): Traditional SMEs – following the low growth development pathway, these concerns generally have few, if any, growth aspirations. They principally exist to provide their owner-managers with a source of employment and income. Furthermore, they are frequently operated in a manner consistent with the life-style aspirations of their ownermanagers. The McMahon (forthcoming) study suggests that after approximately 15 years such SMEs would have fewer than 20 employees, sales less than $3 million per annum, total assets below $2 million, little or no employment growth, and sales growth up to 5 per cent per annum. Capped growth SMEs – following the moderate growth development pathway, these concerns generally have modest growth aspirations. Bounds to growth could be externally imposed by the nature of their competitive environment; or may be intrinsic given the nature of their operations. Frequently though, growth is deliberately capped by owner-managers to a rate that limits dependence upon external financing – thus minimising surrender of control and accountability obligations this support would normally bring. The McMahon (forthcoming) study suggests that after approximately 15 years such SMEs would have fewer than 100 employees, sales around $10 million 6 per annum, total assets less than $10 million, employment growth up to 3 per cent per annum, and sales growth as much as 10 per cent per annum. Entrepreneurial SMEs – following the high growth development pathway, these concerns generally have ambitious growth aspirations. They are most often associated with entrepreneurial aptitude, international outlook, technical and commercial innovation, and other business qualities that could see them eventually become large enterprises. The McMahon (forthcoming) study suggests that after approximately 15 years such SMEs would have over 100 employees, sales around $30 million per annum, total assets more than $20 million, employment growth exceeding 5 per cent per annum, and sales growth greater than 10 per cent per annum. The fact that these common SME configurations are recognised in the research lends further plausibility to the empirically-based development taxonomy derived. SME Growth, Exporting and Innovation Predating the recent phenomenon of ‘born global’ SMEs, Welch (1977, p. 3) places the internationalisation of businesses within the context of their broader growth aspirations as follows: The international growth of the firm should be viewed as part of the overall growth process of the firm, initially growing out of its domestic activities but remaining inextricably entwined with them. Internationalisation undoubtedly adds a new dimension to the firm’s total operations but it builds on and interacts with total expansion. As a consequence, international decisions are not simply explained by what happens within foreign markets. Subsequent support for this position is provided by Vozikis & Mescon (1988), Kamath et al. (1987) and McAuley (1988) amongst other researchers. A recent report from the Australian federal government (Department of Foreign Affairs and Trade, 1995, p. 4) points out that: 7 Most SMEs . . . are globalised only to a limited extent, and are best described as being internationalised. An SME exporting to one or two countries in the region would not consider itself to be globalised, although it is clearly engaged in international activity. Another key report on internationalised SME in the same year observes that (Australian Manufacturing Council, 1995, p. 11): Some of the firms . . . export only 8 to 10 per cent of their sales. In the past we have called such exporters opportunistic, arguing that a firm had to export 20 per cent of sales to be considered a strategic or serious exporter. However, this distinction is inappropriate when small-scale exporters produce global products. The learning opportunities gained by being a player (even a small player) in the international economy are the real advantage that exporting provides. This research adopts the latter, broader view on the significance of SME exporting. Innovation refers to creation of commercial value through introduction of new or improved products or processes (Australian Manufacturing Council, 1995). Linkages between SME growth, exporting and innovation constitute a major theme in a number of significant reports to or from governmental/quasi-governmental organisations in Australia during the early to mid-1990s. They include the Emerging Exporters report (McKinsey & Company, 1993), the Small Business Innovation report (Bureau of Industry Economics, 1994), the Wealth of Ideas report (McKinsey & Company, 1994), the Innovation Cycle report (Australian Manufacturing Council, 1995), and the Winning Enterprises report (Department of Foreign Affairs and Trade, 1995). These place a great deal of emphasis, if not complete focus, upon manufacturing SMEs. On the linkage between SME growth and innovation, the Small Business Innovation report (Bureau of Industry Economics, 1994, p. 23) points out that: Innovations may be used as a strategy for growth or survival. Growth may be achieved if innovation enables the expansion of existing markets or entry into new markets. On the other hand, firms may be forced to match the innovation of rivals to maintain market share. 8 On the linkage between SME innovation and exporting, the Emerging Exporters report (McKinsey & Company, 1993) identifies four ways in which technological innovation can help to create competitive advantage in international markets: The initial success of ‘born global’ SMEs is often based on technologically unique products. In order to internationalise their operations, SMEs already established in domestic markets usually need to develop new products tailored to the demands of export markets. Once established in international markets, SMEs need to engage in continuous product and process innovation in order to enjoy on-going success. SMEs exploiting niche overseas markets inevitably face sales ceilings, and so need to innovate in order to enter new niches and continue to grow. The Emerging Exporters report (McKinsey & Company, 1993) indicates that 44 per cent of emerging exporters growing at a rate greater than 15 per cent per annum developed new products for export. Of emerging exporters growing at less than 15 per cent per annum, only 17 per cent developed new products for export. The Emerging Exporters report (McKinsey & Company, 1993) also reveals that most emerging exporters develop new technologies and products in-house (that is, undertake their own research and development). An appropriate perspective on this finding is provided by the Small Business Innovation report (Bureau of Industry Economics, 1994, p. 5): R&D is only one of the ways in which firms can acquire the knowledge required to introduce a technical innovation – information can also be obtained from parent firms, by licensing, by purchase, by accessing publicly available information and so on – but it is the route on which innovation policy in Australia and many other countries focus. The Wealth of Ideas report (McKinsey & Company, 1994) and the Winning Enterprises report (Department of Foreign Affairs and Trade, 1995) both strongly reflect this perspective. 9 Finally, the case for a linkage between SME exporting and innovation is put most persuasively by the Innovation Cycle report (Australian Manufacturing Council, 1995, pp. 56) as follows: Of all the variables that we examined, in terms of both practices and outcomes, that which most differentiates innovative firms from the rest is exporting. Innovative firms are much more likely to export. They can compete against the rest of the world. Exporting returns to firms more than just the value of their sales. It puts them out in the most competitive market places where there is the greatest pressure to improve performance. The pressure on firms to improve is balanced by their enhanced capacity to deliver. This capacity is provided by access to new customers, a wide range of information and exposure to new ideas. Some of the additional capacity is provided by the linkage relationships that exporting develops. The Wealth of Ideas report (McKinsey & Company, 1994) and the Winning Enterprises report (Department of Foreign Affairs and Trade, 1995) explain the significance of the ‘linkage relationships’ alluded to at the close of this quotation. They most often entail close collaboration with technologically demanding foreign customers that may result in sharing of the costs and risks of innovation. They may also extend to privileged access to foreign research and development resources and technological infrastructure. The central message for the current research from the reports reviewed above is clear and unambiguous. However, notwithstanding their many merits, none of these reports actually demonstrate, using appropriate statistical analysis of large-scale representative samples, the existence, persistence and usefulness of the conjectured linkages between SME growth, exporting and innovation amongst Australian manufacturers. Thus, the two broad hypotheses to be tested in the remainder of this paper are as follows: H1: That linkages between growth, exporting and innovation exist amongst Australian manufacturing SMEs, that they are statistically significant, and that they persist over time. 10 H2: That knowledge of exporting and innovation behaviour amongst Australian manufacturing SMEs can be reliably used to anticipate their growth potential for (inter alia) research and government support purposes. RESEARCH METHOD The panel data employed in this research are drawn from the BLS conducted by the Australian Bureau of Statistics (ABS) on behalf of the federal government over the four financial years 1994-95 to 1997-98. Costing in excess of $4 million, the BLS was designed to provide information on growth and performance of Australian employing businesses. The ABS Business Register was used as the population frame for the survey, with approximately 13,000 business units being selected for inclusion in the 1994-95 mailing of questionnaires. For the 1995-96 survey, a sub-sample of the original selections for 1994-95 was chosen, and this was supplemented with a sample of new business units added to the Business Register during 1995-96. The sample for the 1996-97 survey was again in two parts. The first formed the longitudinal or continuing part of the sample, comprising all those remaining live businesses from the 1995-96 survey. The second part comprised a sample of new business units added to the Business Register during 1996-97. A similar procedure was followed for the 1997-98 survey. Approximately 6,400 business units were surveyed in each of 1995-96, 1996-97 and 1997-98. The BLS did not employ completely random samples. The original population (for 1994-95) was stratified by industry and business size, with equal probability sampling methods being employed within strata. Further stratification by innovation, exporting and growth status took place for the 1995-96 survey. Data collection in the BLS was achieved through self-administered, structured questionnaires containing essentially closed questions. Copies of the questionnaires used in each of the four annual collections can be obtained from the ABS. Because information collected in the BLS was sought under the authority of the Census and Statistics Act 1905, 11 and thus provision of appropriate responses to the mailed questionnaires could be legally enforced by the Australian Statistician, response rates were very high by conventional research standards – typically exceeding 90 per cent. The specific BLS data used in this study are included in a Confidentialised Unit Record File (CURF) released by the ABS on CD-ROM in December, 1999. This CURF contains data on 9,731 business units employing fewer than 200 persons – broadly representing SMEs in the Australian context. Restricted industrial classification detail, no geographical indicators, presentation of enterprise age in ranges, and omission of certain data items obtained in the BLS all help to maintain the confidentiality of unit records. This research is concerned only with manufacturing SMEs in the BLS CURF, representing approximately 35 per cent of businesses in the file. Additional focus is provided by considering only manufacturing SMEs legally organised as proprietary companies. Approximately 71 per cent of manufacturing SMEs in the file are proprietary companies. Variables used in this research are either categorical in nature or, if metric, have irregular distributional properties. Thus, non-parametric/distribution free techniques of statistical analysis are employed exclusively. RESEARCH FINDINGS Existence of SME Growth, Exporting and Innovation Linkages The exporting and innovation behaviour of manufacturing SMEs in the study are initially captured in a series of dichotomous variables as follows: Exporter status – whether or not the businesses exported goods in each of the four years in the longitudinal panel. Innovator status – whether or not the businesses developed or introduced any new or substantially changed products or services in each of the four years in the longitudinal panel. 12 Exporter and innovator status – whether or not the businesses both exported goods and developed or introduced any new or substantially changed products or services in each of the four years in the longitudinal panel. Frequency distributions for these variables across low, moderate and high growth development pathways identified amongst these manufacturing SMEs by McMahon (forthcoming) are presented in Table 1. Inspection of Table 1 reveals that approximately 30 to 40 per cent of SMEs on the low growth pathway report export sales. This compares with around 55 to 60 per cent of businesses on the moderate growth pathway, and roughly 60 to 70 per cent of those on the high growth pathway. Hence, exporting appears to be more prevalent amongst higher growth concerns. Approximately 20 to 40 per cent of SMEs on the low growth pathway appear to be innovators. This compares with around 30 to 50 per cent of businesses on the moderate growth pathway, and roughly 40 to 60 per cent of those on the high growth pathway. Here, innovation appears to be more prevalent amongst higher growth concerns. Finally, approximately 10 to 20 per cent of SMEs on the low growth pathway report being both exporters and innovators. This compares with around 20 to 30 per cent of businesses on the moderate growth pathway, and roughly 30 to 40 per cent of those on the high growth pathway. Thus, being both an exporter and an innovator appears to be more prevalent amongst higher growth concerns. Further evidence for the existence of linkages between SME growth, exporting and innovation in the study sample is provided by examining a series of metric variables reflecting the extent of exporting or innovation activities as follows: Export intensity – the percentage of total sales accounted for by export sales in each of the four years in the longitudinal panel. 13 Research and development commitment – expenditure on research and development expressed as a percentage of total sales in each of the four years in the longitudinal panel. Innovation commitment – total expenditure on innovation (including research and development, training staff, acquisition of patents/trademarks/licences, tooling up, industrial engineering, manufacturing start-up, and marketing new products) expressed as a percentage of total sales in three of the four years in the longitudinal panel (total innovation expenditure is not available for 1995-96). Certain statistics (mean, standard deviation and coefficient of variation) for these variables across low, moderate and high growth SME development pathways are presented in Table 2. Inspection of Table 2 reveals that export intensity is approximately 5 per cent for SMEs on the low growth pathway. This compares with around 8 per cent for businesses on the moderate growth pathway, and roughly 9 per cent for those on the high growth pathway. Hence, export intensity appears to be higher amongst higher growth concerns. Notice also that variability in export intensity seems to decrease for higher growth concerns. Research and development commitment is approximately 1.2 per cent for SMEs on the low growth pathway. This compares with around 0.7 per cent for businesses on the moderate growth pathway, and roughly 0.6 per cent for those on the high growth pathway. Here, the relative commitment to research and development appears to be lower amongst higher growth concerns. Of course, the absolute commitment to research and development in dollar terms is likely to be higher for higher growth concerns. Variability in relative commitment to research and development seems to decrease for higher growth concerns. Finally, innovation commitment is approximately 1.9 per cent for SMEs on the low growth pathway. This compares with around 1.5 per cent for businesses on the moderate growth pathway, and roughly 1.1 per cent for those on the high growth pathway. Thus, the relative commitment to 14 innovation appears to be lower amongst higher growth concerns. As with research and development, the absolute commitment to innovation in dollar terms is likely to be higher for higher growth concerns. Variability in relative commitment to innovation seems to decrease for higher growth concerns. Statistical Significance of SME Growth, Exporting and Innovation Linkages Having established that linkages do seem to exist between growth, exporting and innovation amongst manufacturing SMEs in the study sample, it is now appropriate to determine whether these linkages appear to be statistically significant. Initially, the statistical significance of relationships between the exporter, innovator, and exporter and innovator status variables across low, moderate and high growth development pathways can be examined. This is done using 2 tests in Table 3. It is clear that highly statistically significant relationships do exist between the exporter status and innovator status variables in each of the four years in the longitudinal panel. The evidence is that exporting SMEs are more than likely to also be innovators. Focusing on the exporting linkage with SME growth, over the four years in the longitudinal panel highly statistically significant relationships exist between the exporter status variables and the variable indicating SME development pathways. Higher growth manufacturing SMEs are more than likely to also be exporters. The statistical significance of the relationship between the innovator status variables and the variable indicating SME development pathway is more equivocal. In two of the four years in the longitudinal panel the relationship is not significant; whereas in the other two years the relationship is significant. Amongst other things, this suggests that innovation is a less significant and/or less consistent concomitant with SME growth than exporting. Finally, the relationship between the exporter and innovator status variables and the variable indicating SME development pathways is 15 highly statistically significant in all four years in the longitudinal panel. Higher growth manufacturing SMEs are more than likely to also be both exporters and innovators. Now, a series of Kruskal-Wallis one-way analysis of variance tests involving variables reflecting the extent of exporting or innovation activities can be examined. These are presented in Table 4. First, Kruskal-Wallis tests lead to rejection of null hypotheses that median values for research and development commitment do not differ between nonexporters and exporters over the four years in the longitudinal panel. Kruskal-Wallis tests also lead to rejection of null hypotheses that median values for innovation commitment do not differ between non-exporters and exporters over the four years in the longitudinal panel. Both research and development commitment and innovation commitment are found to be higher amongst exporting SMEs. Further Kruskal-Wallis tests lead to rejection of null hypotheses that median values for export intensity do not differ between non-innovators and innovators over the four years in the longitudinal panel. Export intensity is found to be higher amongst innovating SMEs. Together, the three series of tests support the earlier finding of a highly statistically significant relationship between SME exporting and innovation. Focusing on the exporting linkage with SME growth, Kruskal-Wallis tests lead to rejection of null hypotheses that median values for export intensity do not differ between development pathways. Export intensity is found to be higher amongst higher growth manufacturing SMEs. As before, the statistical significance of the relationship between innovation and SME development pathway is more equivocal. In two of the four years in the longitudinal panel, Kruskal-Wallis tests lead to acceptance of null hypotheses that median values for research and development commitment do not differ between development pathways. Furthermore, in two of three years in the longitudinal panel, Kruskal-Wallis tests lead to acceptance of null hypotheses that median values for innovation commitment do not 16 differ between development pathways. Again, the suggestion is that innovation is a less significant and/or less consistent concomitant with SME growth than exporting. Persistence of SME Growth, Exporting and Innovation Linkages It is now appropriate to ascertain whether linkages that exist between growth, exporting and innovation amongst manufacturing SMEs in the study sample appear to persist over time. If they are found to be persistent, this lends further support to the contention that such linkages are sufficiently important to warrant further attention from researchers and policymakers in the field. While the data and analyses presented so far in the paper may suggest the persistence of linkages between growth, exporting and innovation amongst the manufacturing SMEs studied, this impression only applies to the businesses in aggregate. What is sought is an indication of whether there seems to be continuity of the linkages at the individual business level over the four years of the longitudinal panel. This is gained by examining the results of a series of Friedman two-way analysis of variance tests presented in Table 5. Consider first the stability of the exporter status variables over time across low, moderate and high growth development pathways. For both moderate and high growth SMEs, it is not possible to reject null hypotheses that exporter status is the same over the four years of the longitudinal panel. It is, however, possible to reject such a null hypothesis for low growth SMEs. The suggestion here is that individual businesses following the low growth development pathway tend to change exporter status over time; whereas this is not the case for manufacturing SMEs on the moderate and high growth development pathways. This is consistent with the earlier claim that many smaller SMEs are simply opportunistic exporters. It would appear that more regular exporting activity may be required in order to follow higher growth development pathways. 17 Now consider the stability of the innovator status variables over time across low, moderate and high growth development pathways. For all three pathways, it is possible to reject null hypotheses that innovator status is the same over the four years of the longitudinal panel. The suggestion here is that, regardless of their growth aspirations, individual businesses tend to change innovator status over time. This reinforces the earlier impression that innovation is a less significant and/or less consistent concomitant with SME growth than exporting. The stability of the exporter and innovator status variables over time across low, moderate and high growth development pathways is only marginally greater. Only for high growth SMEs is it not possible to reject the null hypothesis that exporter and innovator status is the same over the four years of the longitudinal panel. Thus, it would appear that more regular exporting and innovation activity may be required in order to follow the high growth development pathway. It remains to examine the stability over time of variables reflecting the extent of exporting or innovation activities. For all three development pathways, it is not possible to reject null hypotheses that export intensity is the same over the four years of the longitudinal panel. The suggestion here is that, regardless of their growth aspirations, individual businesses tend not to change their relative dependence on export versus domestic sales over time. In other words, the relative importance of exporting does not seem to change regardless of the development pathway being followed. Could it be that manufacturing SMEs in the study sample undertake a certain level of exporting consistent with their growth aspirations, and then simply maintain that degree of commitment to international activity? Perhaps there is a ‘comfort zone’ in terms of relative dependence on export sales that SMEs choose to be in. If so, policy initiatives to increase the number of SMEs exporting are more likely to be effective than efforts to increase the level of exporting amongst those concerns that have already taken a measured plunge into international business. 18 Consider now the stability of the research and development commitment variables over time across low, moderate and high growth development pathways. For both low and moderate growth SMEs, it is possible to reject null hypotheses that research and development commitment is the same over the four years of the longitudinal panel. However, it is not possible to reject such a null hypothesis for high growth SMEs. A more consistent commitment to research and development may be required in order to follow the high growth development pathway. Finally, consider the stability of the innovation commitment variables over time across low, moderate and high growth development pathways. For all three development pathways, it is possible to reject null hypotheses that innovation commitment is the same over the four years of the longitudinal panel. The suggestion here is that, regardless of their growth aspirations, individual businesses tend to change their commitment to innovation over time. Together, the findings for research and development commitment and for innovation commitment add to equivocation about the significance and/or persistence of the linkage between manufacturing SME growth and innovation Anticipating SME Growth Potential from Exporting and Innovation Behaviour It remains to ascertain whether knowledge of linkages between growth, exporting and innovation amongst manufacturing SMEs in the study sample can be reliably used to anticipate the growth potential of manufacturing SMEs for, say, research and government support purposes. Because of the categorical nature of the dependent variable (SME development pathway with three values for low growth, moderate growth and high growth), the appropriate analytical procedure for anticipating the growth potential of the manufacturing concerns investigated is logistic regression. The assumptions underlying logistic regression are undemanding and its use with the irregularly distributed data available to the present study is entirely appropriate (Aldrich & Nelson, 1984). For research and government support 19 purposes, the main consideration in evaluating models of this kind is the minimisation of Type 2 errors – that is, avoiding identification of a business as a non-growth SME when, in fact, it is a growth SME. The possible economic costs of failing to recognise a high growth SME in need of support require no elaboration. In these circumstances, the most important means for assessing the goodness-of-fit of a logistic regression model is through its classification success. Classification is usually achieved by assigning each case to the dependent variable category for which the estimated probability exceeds 0.5. Key statistics include the proportion of classifications that are correct by dependent variable category. A comparison of the overall classification success with an arbitrary benchmark of being 50 per cent correct (achievable through the toss of a fair coin) is also of interest (Peel & Peel, 1988). Focusing on dichotomous logistic regression, Hosmer & Lemeshow (1989, p. 147) note that: Classification is sensitive to the relative sizes of the two component groups and will always favour classification into the larger group, a fact that is also independent of the fit of the model. Ohlson (1980) also observes that logistic regression is not a technique designed to find an optimal trade-off between Type 1 and Type 2 errors. Preliminary model building that attempted to correctly classify manufacturing businesses over the three SME development pathways (that is, with a trichotomous dependent variable) using exporting and innovation measures in the BLS data as independent variables proved to be most discouraging. In the main, classification success was markedly worse than the toss of a fair coin. For this reason, binary logistic regression with a dichotomous dependent variable distinguishing high growth SMEs from other concerns (low and moderate growth SMEs) was trialled. A justification for this approach is that it is high growth SMEs that are of greatest interest. Their identification, often referred to in the relevant literature as 20 ‘picking winners’, has been a major preoccupation of researchers and policy-makers worldwide (Turok, 1991). Findings on classification success for certain binary logistic regression models are presented in Table 6. It can be seen that logistic regression models using exporter, innovator, and exporter and innovator status over the four years of the longitudinal panel as independent variables, either separately or together, produce zero classification success for the 39 high growth SMEs in the study sample. Furthermore, even when the status variables are supplemented with variables reflecting export intensity, research and development commitment, and innovation commitment, the classification success for high growth SMEs remains zero. Although they had already been employed in derivation of the export intensity, research and development commitment, and innovation commitment variables, it was decided to also trial as independent variables the absolute dollar values for export sales, research and development expenditure, and innovation expenditure over the four years of the longitudinal panel. Such variables obviously introduce an element of enterprise size or scale to the modelling. Using these variables together with exporter, innovator, and exporter and innovator status produces a classification success of just 15.4 per cent for high growth SMEs. However, when all export and innovation behaviour measures that have been identified are included as independent variables a more encouraging classification success of 74.4 per cent is produced for high growth SMEs. Clearly, this is markedly better than a benchmark of being 50 per cent correct. Stepwise elimination of non-statistically significant independent variables inevitably resulted in poorer classification success. Thus, all export and innovation behaviour measures for all four years of the BLS panel are required as independent variables to produce maximum classification success for the high growth SMEs studied. 21 A major limitation of the predictive modelling undertaken should be acknowledged. Classification success for a logistic regression model can be determined for the original sample from which the model has been developed, for a sample withheld from the original data set, or for an entirely new set of data. On the need for validation of a logistic regression model with data other than that from which it has been developed, Hosmer & Lemeshow (1989, p. 171) comment as follows: The reason for considering this type of assessment of model performance is that the fitted model always performs in an optimistic manner on the developmental data set. In the present exploratory context, classification success has been evaluated only for the original modelling sample. A hold-out sample was not practical since there are only 39 high growth SMEs in the on-going longitudinal panel of 871 manufacturing SMEs. Furthermore, no external validation has been undertaken using data from an entirely separate source because, for cost and other reasons, such data are not available to the researcher. CONCLUSIONS Based on evidence presented, the first broad hypothesis for testing in this research has substantial but not complete support. Linkages between growth, exporting and innovation do appear to exist amongst Australian manufacturing SMEs in the BLS panel. In the main, these linkages are statistically significant; and some of them do seem to persist over time. There is a strong suggestion that innovation is a less significant and/or less consistent concomitant with SME growth than exporting. An explanation for this could be that innovation is an intrinsically cyclical activity, with periods of innovative effort being followed by some time for commercial consolidation. Eventually, technological change, increased competitive pressure or some such trigger may necessitate another period of innovative effort to ensure continued success. In contrast, once initiated, export activity is much more likely to be an ongoing and regular commitment to overseas customers secured. 22 Based on evidence presented, support is less compelling for the second broad hypothesis for testing in this research. It would appear that, using export and innovation behaviour variables available in the BLS data, success in anticipating the growth potential of the manufacturing SMEs studied can, at best, be described as modest. Using a cumbersome logistic regression model with all export and innovation behaviour measures for all four years of the longitudinal panel as independent variables, the best achievable classification success for high growth SMEs is around 75 per cent. Of the 39 high growth SMEs in the data set, 10 are not identified as such. From the viewpoint of researchers and policy-makers, the economic consequences of this degree of failure could, of course, be non-trivial. Thus, it would seem to have been demonstrated that, while export and innovation behaviour are useful pointers to growth potential, there is a need to look more widely for growth correlates if the holy grail of being able to reliably ‘pick winners’ amongst manufacturing SMEs is ever to be realised. 23 REFERENCES Aldrich, J.H. & F.D. Nelson. 1984. Linear Probability, Logit, and Probit Models. Beverly Hills, California: Sage Publications. Australian Manufacturing Council. 1995. The Innovation Cycle: Practical Tips from Innovative Firms. Melbourne, Victoria: Australian Manufacturing Council. Bureau of Industry Economics. 1994. A Report of Small Business Innovation. Small Business Review 1993. Canberra, Australian Capital Territory: Australian Government Publishing Service. Department of Foreign Affairs and Trade. 1995. Winning Enterprises: How Australia’s Small and Medium Enterprises Compete in Global Markets. Canberra, Australian Capital Territory: Australian Government Publishing Service. Hanks, S.H., C.J. Watson, E. Jansen & G.N. Chandler. 1993. Tightening the life-cycle construct: a taxonomic study of growth stage configurations in high-technology organizations. Entrepreneurship Theory and Practice, 18(2): 5-29. Hosmer, D.W. & S. Lemeshow. 1989. Applied Logistic Regression. New York, New York: John Wiley & Sons. Kamath, S., P. Rosson, D. Patton & M. Brooks. 1987. Research on success in exporting: past, present, and future. In P.J. Rosson & S.D. Reid, editors, Managing Export Entry and Expansion. New York, New York: Praeger Publishers. McAuley, A. 1988. The role of internationalisation in the growth of firms. Paper presented at 11th National Small Firms Policy and Research Conference, Scottish Enterprise Foundation, University of Stirling, Scotland. McKinsey & Company. 1993. Emerging Exporters: Australia’s High Value-Added Manufacturing Exporters. Melbourne, Victoria: Australian Manufacturing Council. 24 McKinsey & Company. 1994. The Wealth of Ideas: How Linkages Help Sustain Innovation and Growth. Melbourne, Victoria: Australian Manufacturing Council. McMahon, R.G.P. Forthcoming. Deriving an empirical development taxonomy for manufacturing SMEs using data from Australia’s business longitudinal survey. McMahon, R.G.P., S. Holmes, P.J. Hutchinson & D.M. Forsaith. 1993. Small Enterprise Financial Management: Theory and Practice. Sydney, New South Wales: Harcourt Brace. Ohlson, J.A. 1980. Financial ratios and the probabilistic prediction of bankruptcy. Journal of Accounting Research, 18(1): 109-131. Peel, M.J. & D.A. Peel. 1988. A multilogit approach to predicting corporate failure – some evidence for the UK corporate sector. Omega: International Journal of Management Science, 16(4): 309-318. Turok, I. 1991. Which small firms grow?. In L.G. Davies & A.A. Gibb, editors, Recent Research in Entrepreneurship. England: Avebury. Vozikis, G.S. & W. Mescon. 1988. Stages of development and stages of the exporting process in a small business context. In R.J. Rudd, W.T. Greenwood & F.W. Becker, editors, Small Business in a Regulated Economy: Issues and Policy Indications. New York, New York: Quorom Books. 25 Table 1: FREQUENCY DISTRIBUTIONS FOR GROWTH, EXPORTING AND INNOVATING SMEs Innovatorsb Exporters & Innovators Exportersa Innovatorsb Exporters & Innovators Exportersa Innovatorsb Exporters & Innovators High growth SMEs (n=39) Exportersa Moderate growth SMEs (n=203) 232 242 104 215 193 82 199 131 71 207 112 63 36.9% 38.5% 16.5% 34.2% 30.7% 13.0% 31.6% 20.8% 11.3% 32.9% 17.8% 10.0% 112 109 65 114 82 56 117 51 37 122 61 45 55.2% 53.7% 32.0% 56.2% 40.4% 27.6% 57.6% 25.1% 18.2% 60.1% 30.0% 22.2% 22 26 17 27 16 13 24 12 10 24 15 12 56.4% 66.7% 43.6% 69.2% 41.0% 33.3% 61.5% 30.8% 25.6% 61.5% 38.5% 30.8% a Includes both innovating and non-innovating SMEs. b 1997-98 Exporters & Innovators Low growth SMEs (n=629) 1996-97 Innovatorsb Growth Pathway 1995-96 Exportersa 1994-95 Includes both exporting and non-exporting SMEs. 26 Table 2: EXPORT INTENSITY AND R&D/INNOVATION COMMITMENT STATISTICS FOR SME GROWTH PATHWAYSa a R&D Commitmentc Innovation Commitmentd Export Intensityb R&D Commitmentc Innovation Commitmentd Export Intensityb R&D Commitmentc Innovation Commitmentd High growth SMEs (n=39) Export Intensityb Moderate growth SMEs (n=203) 1997-98 Innovation Commitmentd Low growth SMEs (n=629) 1996-97 R&D Commitmentc Growth Pathway 1995-96 Export Intensityb 1994-95 5.0 1.5 3.1 5.3 1.2 n.a. 5.2 1.0 1.4 5.0 1.0 1.1 14.5 6.5 11.4 14.6 6.5 n.a. 14.5 6.4 7.3 14.0 5.3 4.8 2.9 4.3 3.7 2.8 5.4 n.a. 2.8 6.4 5.2 2.8 5.3 4.4 7.4 0.8 2.3 8.1 0.8 n.a. 8.7 0.5 0.7 8.5 0.6 1.4 17.3 2.4 5.4 16.9 2.7 n.a. 18.1 1.7 1.9 17.6 2.7 6.0 2.3 3.0 2.3 2.1 3.4 n.a. 2.1 3.4 2.7 2.1 4.5 4.3 8.5 0.8 1.6 10.6 0.6 n.a. 8.9 0.5 0.8 9.3 0.6 0.8 17.6 1.5 2.3 19.8 1.5 n.a. 18.5 1.6 1.8 20.8 1.7 2.0 2.1 1.9 1.4 1.9 2.5 n.a. 2.1 3.2 2.3 2.2 2.8 2.5 Statistics are mean, standard deviation and coefficient of variation. Export intensity = (Annual export sales/Annual total sales) x 100%. c R&D commitment = (Annual R&D expenditure/Annual total sales) x 100%. d Innovation commitment = (Annual innovation expenditure/Annual total sales) x 100%. b 27 Table 3: 2 TESTS FOR GROWTH, EXPORTING AND INNOVATING SMEs Test Values Test Variables Test Details Exporter status vs Innovator status 2 statistic df Significance Growth pathway vs Exporter status 2 statistic df Significance Growth pathway vs Innovator status 2 statistic df Significance Growth pathway vs Exporter & Innovator status 2 statistic df Significance 1994-95 1995-96 1996-97 1997-98 14.604 21.950 49.794 54.008 1 1 1 1 0.000 0.000 0.000 0.000 24.539 44.256 52.275 54.548 2 2 2 2 0.000 0.000 0.000 0.000 23.577 7.569 3.339 20.463 2 2 2 2 0.000 0.023 0.188 0.000 33.932 29.961 11.407 29.000 2 2 2 2 0.000 0.000 0.003 0.000 28 Table 4: KRUSKAL-WALLIS TESTS FOR EXPORT INTENSITY AND R&D/INNOVATION COMMITMENT Test Values 1994-95 1995-96 1996-97 1997-98 29.415 60.476 50.970 61.166 1 1 1 1 0.000 0.000 0.000 0.000 12.573 n.a. 53.713 50.921 1 n.a. 1 1 0.000 n.a. 0.000 0.000 17.220 23.171 51.325 51.427 1 1 1 1 0.000 0.000 0.000 0.000 23.739 40.324 48.921 49.736 2 2 2 2 Significance 0.000 0.000 0.000 0.000 Kruskal-Wallis statistic 7.619 21.783 1.376 11.357 2 2 2 2 Significance 0.022 0.000 0.503 0.003 Kruskal-Wallis statistic 7.331 n.a. 2.588 16.477 2 n.a. 2 2 0.026 n.a. 0.274 0.000 Test Variables Test Details Exporter status vs R&D commitment Kruskal-Wallis statistic df Significance Exporter status vs Innovation commitment Kruskal-Wallis statistic df Significance Innovator status vs Export intensity Kruskal-Wallis statistic df Significance Growth pathway vs Export intensity Growth pathway vs R&D commitment Growth pathway vs Innovation commitment Kruskal-Wallis statistic df df df Significance 29 Table 5: FRIEDMAN TESTS FOR GROWTH, EXPORTING AND INNOVATING SMEs Test Values Export/Innovation Measure Test Details Exporter status Low growth SMEs (n=629) Innovator status Exporter & Innovator status Export intensity R&D commitment df Significance 17.947 3 0.000 Moderate growth SMEs (n=203) 3.721 3 0.293 High growth SMEs (n=39) 4.636 3 0.200 122.745 3 0.000 Moderate growth SMEs (n=203) 55.238 3 0.000 High growth SMEs (n=39) 15.276 3 0.002 Low growth SMEs (n=629) 24.675 3 0.000 Moderate growth SMEs (n=203) 17.248 3 0.001 High growth SMEs (n=39) 4.216 3 0.239 Low growth SMEs (n=629) 1.256 3 0.740 Moderate growth SMEs (n=203) 6.693 3 0.082 High growth SMEs (n=39) 1.879 3 0.598 Low growth SMEs (n=629) 84.643 3 0.000 Moderate growth SMEs (n=203) 37.241 3 0.000 4.176 3 0.243 115.282 2 0.000 51.029 2 0.000 9.556 2 0.008 Low growth SMEs (n=629) High growth SMEs (n=39) Innovation commitment Statistic Low growth SMEs (n=629) Moderate growth SMEs (n=203) High growth SMEs (n=39) 30 Table 6: CLASSIFICATION SUCCESS FOR LOGISTIC REGRESSION MODELS (n=871) Predicted Classification Independent Variable(s) (from 1994-95 to 1997-98) Exporter status Observed Classification Not high growth High growth Innovator status Not high growth High growth Exporter & innovator status Not high growth High growth Exporter status + Innovator status Exporter status + Innovator status + Exporter & innovator status Exporter/innovator status variables (as above) + Export intensity and R&D/innovation commitment Exporter/innovator status variables + Annual export sales, R&D expenditure and innovation expenditure Exporter/innovator status variables + Export intensity and R&D/innovation commitment + Annual export sales, R&D expenditure and innovation expenditure Not high growth High growth Not high growth High growth Not high growth High growth Not high growth High growth Not high growth High growth Not High Growth High Growth Per Cent Correct 832 0 100.0 39 0 0.0 Overall per cent correct 95.5 832 0 100.0 39 0 0.0 Overall per cent correct 95.5 832 0 100.0 39 0 0.0 Overall per cent correct 95.5 832 0 100.0 39 0 0.0 Overall per cent correct 95.5 832 0 100.0 39 0 0.0 Overall per cent correct 95.5 832 0 100.0 39 0 0.0 Overall per cent correct 95.5 829 3 99.6 33 6 15.4 Overall per cent correct 95.9 830 2 99.8 10 29 74.4 Overall per cent correct 98.6