FirstPreviousNextLast

Canadian SME Exporters

4. Empirical Findings

4.1 Descriptive Findings

In 2004, 8 percent of Canadian SMEs exported goods or services. Exporters (both INVs and established exporters) derived an average of 33 percent of sales revenues from exports. Approximately half (52.3 percent) of exporters report export revenue accounts for less than 25 percent of total sales; 15 percent report export revenue accounts for 25 to 49 percent of revenue, while a third (32.8 percent) report export revenue in excess of 50 percent of sales.

It was also noted that a high proportion of exporters derived a relatively small proportion of revenues from exports: 44 percent of exporters reported export intensities (percent of sales from exports) of less than ten percent. Arguably, some of the firms that exhibited low export intensity may not truly be active exporters: cross border sales may have been one-off occasions, etc. Accordingly, an alternative (and more strict) definition of SME exporter was used by defining as "intensive exporters" those firms that derived more than 25 percent of sales revenues from export sales. Of the 368 INVs in the sample, 194 were intensive exporters and 371 of the 865 established exporters were "intensive" exporters according to this definition.

4.2 Owner and Firm Attributes

Table 3 presents a comparison of owner (Table 3A) and firm (Table 3B) attributes of each of the categories of firms of interest: international new ventures, domestic new ventures, established exporters and established non-exporters. Table 3 (A and B) reveals that the SMEs investigated here are indeed small, with an average of four employees and average annual revenues of less than $600,000. Less than 40 percent of owners expressed an intention to grow their firms in the next two years. Firms operate across sectors, with vast majority being in the services sectors; approximately five percent of firms are in the manufacturing firms with a similar proportion of firms in knowledge-based industries. Approximately one-third (28.7 percent) of Canadian SMEs report some expenditure for R&D and less than 5 percent (4.2 percent) of SMEs are deemed to be "innovators" (defined as R&D expenditures in excess of 25 percent of total investment).

4.2.1 Comparaison of Exporters and Domestic-based Firms

Univariate Student t-tests and Wilcoxon rank sum tests were conducted to identify statistically significant differences between exporters and non-exporters, INVs and domestic-based new ventures (DNVs), and between established exporters (EEs) and established domestic firms. Table 3A reveals that, overall, the owners of exporters and of non-exporter firms differ in very few respects. Contrary to stage theory, owners of exporter firms do not differ to a statistically significant in terms of age or years of management experience. However, they do differ significantly on three important dimensions.

1. Exporters and non-exporters (both new enterprises as well as established businesses) differ significantly (p-value < 0.000) in terms of their respective growth intentions. Owners of exporter firms are much more likely to profess an intention to seek the growth of their businesses than are owners of non-exporters.
   
   
2. Exporters and non-exporters differ significantly in terms of the gender composition of the ownership teams (p-value < 0.000). As noted in previous research, firms owned by a majority of females are less likely to be exporter firms than those owned by a majority of males. Among established firms, this difference is especially striking. Among established domestic firms, 16.1 percent (15.1 percent + 1.0 percent ) are majority owned by females. Among established exporter firms only 7.5 percent are majority owned by females. Hence, majority female-owned firms are less than half as likely to export as firms where men comprise the majority of the ownership teams, an observation that bears further analysis.
   
   
3. Finally, among enterprises that began trading since 2001, recent immigrants were significantly more likely to be the primary owners of exporter firms than of non-exporter businesses (p-value < 0.001).

Where Table 3A compared the attributes of owners across exporters and non-exporters, Table 3B compares the attributes of the businesses themselves. Table 3B shows that exporter firms are significantly larger than non-exporters in terms of number of employees (p-value < 0.000) and appear to be larger in terms of financial measures such as total revenues and total assets.^{Footnote 15} Exporter firms are significantly more likely to invest in R&D and to be situated in urban locations (p-values < 0.000). As expected, the sectoral distribution of exporters and non-exporter firms differs significantly with exporters being more likely to be in the manufacturing, wholesale and retail, professional services and knowledge-based sectors. Exporter firms were more likely to use debt financing; and more likely to have sought external capital and bank loans (p-value < 0.000).

While the univariate comparisons reported in Table 3 (A and B) are supportive of several of study propositions advanced here, they are not necessarily conclusive. This is because several firm and owner attributes are correlated among themselves. By way of one example, it was noted from Table 3A that female-owned firms appeared less likely to export than firms owned by men; however, female-owned firms are also concentrated in sectors (such as personal services) where exporting is less likely; likewise, female-owned firms also tend to be smaller than those owned by men and smaller firms are also less likely to export. Consequently, it is not yet clear whether the gender differences in export propensity are attributable to differences in firm size and sector or whether gender differences are additional to these systemic factors. Likewise, it is arguable that size and sector are themselves interdependent. To disentangle these potentially confounding effects, further analysis is necessary. To do so, a multivariate model of export propensity, one that simultaneously allows for the various potential factors must be estimated.

4.3 Multivariate Modeling of Export Propensity

To investigate factors associated with export propensity, logistic regression was employed as a means of examining the relative influence of firm and owner attributes on the outcome measure corresponding to whether or not a firm exports. Logistic regression is a particularly useful means of representing decision outcomes: it is a technique that makes relatively few statistical assumptions (Hosmer and Lemeshow 1998; Press and Wilson, 1978); one that is robust to the statistical assumptions that are made (Stevens, 1996); and one that closely follows the decision processes made by humans.

Following from the study propositions, owner profile data entered into the models included owner age, gender composition of the management team (majority male ownership; 50-50 ownership; majority female ownership), informal investment (0, 1), growth intention (0, 1), first language of majority owner (English, French and other), Canadian residency status (person resided in Canada for less than 5 years), and management experience (< 5 years; 5-10 years; > 10 years). Firm profile data included firm size (full-time equivalent employees, FTE), a proxy for innovation (R&D expenditures >25%), and capital profile.^{Footnote 16}

Table 4 shows the results of the logistic regression estimates of export propensity. The left-hand panel of Table 4 shows the results of the logistic regression estimation with all of the above variables retained except for the gender composition of the management team. This base model (left hand panel) was statistically significant (p-value of 0.000), with reasonable goodness-of-fit measures (Cox and Snell R² and Nagelekerke R² of 0.093 and 0.222 respectively) with an in-sample prediction accuracy of 92.7 percent. The Hosmer-Lemeshow p-value was 0.807 indicating no significant difference between in-sample predicted and actual outcomes. Overall, these test statistics indicate a more than adequate goodness of fit.

The right-hand side panel of Table 4 presents the estimates of the final logistic regression model of export propensity with non-significant variables in the base model suppressed and with the model expanded to include a categorical variable corresponding to the gender composition of the management team. Addition of the gender-of-ownership measure improved the overall goodness-of-fit estimate to a statistically significant (p-value = 0.037) extent. The results confirm the univariate results in that age of owner, owners' experience, whether the owner was a member of a visible minority, whether the owner was an informal investor, and age of the firm were not significant (p-value > 0.10) determinants of export propensity. The finding that age of firm was not a factor is especially significant because it directly contradicts the stage theory that forms the basis of much public policy.

Significant differences between exporter and non exporter firms include: (as expected) size and sector with larger firms being more likely to export; whether or not the owner of the firm is a recent immigrant (recent immigrants are more likely to be the primary owners of exporter firms) and English-speaking owners are relatively more likely to be primary owners of exporter firms than are owners whose mother tongues are not English (p-value < 0.10). Two variables that were particularly closely correlated with export propensity were owners' growth intentions and investments in R&D. After allowing for the impacts of firm size and sector, firms whose owners had expressed growth intentions were more than twice as likely to be among exporter firms as firms whose owners did not seek growth. Firms that reported investment in R&D were also more than twice as likely to be exporter firms as firms that did not invest in R&D.

Finally, after allowing for all of the above factors, the gender composition of the ownership team is significantly associated with export propensity. The gender composition is statistically associated with export propensity and makes a significant (p-value of 0.037) incremental contribution to the goodness-of-fit of the logistic model after allowing for all of the other factors noted above. The nature of the correlation is such that, given size, sector, etc., firms where the ownership team is primarily women are relatively less likely to be exporters. Having controlled for sector, firm and owner level differences, these results suggest strongly that female-owned firms were still significantly less likely to export. Differences in export propensity were not fully explained by systemic differences in owner and business attributes. The findings lend support to social feminist arguments that the experiences of men and women (exporters) differ and help to illustrate the relevance of feminist thought within well-received theories of SME growth and internationalization.

These results prompted further analysis of two of the findings. The first was the issue of gender of ownership and its link to export propensity. The second issue was the finding that export propensity, in apparent contradiction of the widely promulgated stage theory, did not depend on the age of the firm: that is, additional examination of international new ventures (INVs) seems warranted. These topics are presently examined further.

4.3.1 Gender of Ownership and Export Propensity

Table 5A and 5B present attributes of firms broken down by gender. Three gender categories are employed: firms in which men constitute more than 50 percent of the ownership team; firms in which women constitute more than 50 percent of the ownership team; firms in which ownership is shared equally. Overall, 62.7 percent of firms are majority owned by men, ownership is equally shared in 20.1 percent of businesses, and females are majority owners of 17.2 percent of firms. Table 5A provides breakdowns of attributes of business owners across gender and exporter categories. Table 5B provides breakdowns of attributes of the businesses themselves across gender and exporter groupings. As before, Student t-tests and Wilcoxon tests were conducted to compare attributes across genders by exporter category to identify statistically significant gender differences.

Table 5A shows that, compared with firms that are majority male-owned, majority female-owned firms report less experience, are younger, and more likely to be owned by a member of a visible minority. While the p-values differ, these gender patterns hold for both exporter and non-exporter firms. Table 5B shows that female-owned firms tend to be smaller, and concentrated in service sectors. It is worth noting that female-owned firms did not differ to a statistically significant extent from firms owned by men in terms of investment in R&D. Among non-exporters, women owners were significantly less likely than men to apply for a bank loan or otherwise seek external capital. Among exporter firms, those that were majority female-owned were, if anything, relatively more likely to seek external financing than exporter firms owned by men. Majority female-owned export firms (firms that derive more that 25% of sales from exports) differed from male counterparts across a number of firm and owner attributes, differences that generally mirrored domestic firms.

4.3.2 International New Ventures

In terms of personal attributes, INVs compare to and contrast from other types of firms in several respects.

Attributes of owners

• Owners' age and experience. No significant age differences between owners of INVs and DNVs were observed. Not surprisingly, INV owners are younger than their EE counterparts: 23 per cent of INV owners are less than 40 years old compared with 15 per cent for EEs. INVs and DNVs report significantly less management experience compared to EEs. As would be expected, INVs' business owners have less experience than owners of EEs. This observation also suggests that the nature of experience (rather than tenure) may differ between INV and DNV owners, a possible area for future research.
  
  
• New Canadians. New Canadian residents (immigrants) were significantly more likely to own INVs than DNVs (9.8 percent compared with 3.8 percent; p-value = 0.002). This finding is one that is consistent with network theory expectations.
  
  
• Mother tongue and visible minorities. In accordance with the disproportionate involvement of immigrants among INV owners, statistical differences in 'mother tongue' was also noted between INV and EE owners, with owners of INVs being more likely to report a first language 'other' than English or French (20.1 percent compared to 6.4 percent of EEs, p-value of 0.000); no statistically significant difference was noted between INV and DNVs. Among INVs, 15 percent are owned by individuals from visible minorities, compared to four percent of EEs and ten percent of DNVs (p-value of 0.054).
  
  
• Gender of ownership team. Firms that were majority female-owned firms comprised an unusually high proportion of INVs. Table 3A shows that 27.8 percent of INVs were majority female-owned. This is higher than the 21.6 percent of DNVs that are female-owned and the difference is significant at a p-value of 0.049. Among established firms, however, the reverse is true: firms that were majority female-owned comprise only 7.5 percent of established exporters and 16.1 percent of established non-exporters. This difference is also statistically significant at a p-value of 0.019. While longitudinal data would be required for verification, these data suggest that where a relatively high proportion of female-owned young firms export, the proportion of female-owned exporters may decrease over time. If true, this result might be indicative of gender-specific barriers to exporting.
  
  
• Growth orientation. Table 6 shows that INV owners are indeed significantly more growth oriented than DNVs and EEs. Three-quarters of INV owners intend to grow the firm compared to half of DNVs and 62.3 percent of EEs (p-value = 0.000 and 0.005, respectively).

Attributes of firms

• Size of firm. On average INVs are larger than DNVs (2.22 FTEs compared to 1.39 FTEs, p-value of 0.001). A comparison of total revenue between INVs and DNVs indicates a similar pattern. It is also intriguing that the average INV has amassed labour and revenue typical of the average of all Canadian SMEs, but has accomplished this within a three-year period. These observations refute arguments about a gradual accumulation of labour and financial resources, as postulated by stage theory. It is also interesting that no significant differences in total assets and profitability were identified among INVs, DNVs and EEs. In other words, INVs reach similar level of productivity to EEs in a significantly shorter timeframe, employing less labour input.
  
  
• Sectoral distribution. INVs operate across all industry sectors. Firms in the manufacturing and knowledge-based sectors tend to be more prevalent among exporters than among non-exporters (whether new ventures or established). Among INVs, a relatively high proportion of exporters are professional services firms and, among established exporters, firms in the wholesale and retail sector are relatively more prevalent. These results are important in that they support concerns about the over emphasis of research on manufacturing, technology- and KBI sectors and under emphasis of other sectors (Lefebvre and Lefebvre, 2000).
  
  
• Requirements for financing. Contrary to expectations, INVs do not demonstrate symbiotic relationships that lessen the need for financial resources. On the contrary, INVs were significantly more likely to have sought debt financing compared to DNVs (40.1 percent compared to 21 percent; p-value = 0.088). It is therefore not surprising that INVs are more leveraged (higher debt to asset ratio) than DNVs (p-value = 0.000). There were no significant differences in loan application rates between INVs and EEs (p-value of 0.183), confirming that exporters appear to require additional financing than do domestic firms. Perhaps this is a result of the costs of export market development.

These results empirically demonstrate that particular owner and firm characteristics are associated with INVs. Experience required to operate INVs reflects personal characteristics and international relationships. Capital and tangible assets are both required. The study's findings refute stage theory and are consistent with network theory. Results indicate that although INVs are significantly smaller than EEs (labour and revenue), yet INV size does not preclude performing in the international arena. Thus INVs achieve similar levels of export intensity within three years of inception with a smaller work force, younger and less experienced owners and fewer assets compared to EEs. These observations are consistent with Madsen and Servais (1997) who suggest that INVs accumulate resources in ways which are more efficient than what was anticipated by the stage model. While this observation had been suggested in qualitative research, this study provides large-sample verification.

Source: Canadian SME Exporters (Orser, Spence, Riding and Carrington, 2007)

Source: Canadian SME Exporters (Orser, Spence, Riding and Carrington, 2007)

Source: Canadian SME Exporters (Orser, Spence, Riding and Carrington, 2007)

Source: Canadian SME Exporters (Orser, Spence, Riding and Carrington, 2007)

Source: Canadian SME Exporters (Orser, Spence, Riding and Carrington, 2007)

*Mann-Whitney U Wilcoxon W estimates
Source: Canadian SME Exporters (Orser, Spence, Riding and Carrington, 2007)

*Mann-Whitney U Wilcoxon W estimates
Source: Canadian SME Exporters (Orser, Spence, Riding and Carrington, 2007)

4.4 Examining Size Thresholds

To explore the relationship between threshold size and exporting, two sets of analysis were conducted. The first series of tests repeated, using the FDI data, previous research that relates export propensity to firm size, historically defined as the number of employees. This approach implicitly assumes that export capacity is strictly a function of the quantity of labour input even though classical economic theory posits that the level of output is a function, not only of labour, but of capital stock and managerial capacity as well and that these inputs may be substituted for each other subject to the production technology. Therefore, the second set of tests employ the Cobb-Douglas production function as a framework. The findings of each approach are discussed presently and differences/similarities noted.

4.4.1 Size Thresholds Defined as Number of Employees

The concept of a threshold size implies an S-shape relationship between propensity and size. Therefore, firms in the FDI database were ranked into categories based on size as measured by the number of full-time-equivalent employees. The categories were constructed so that each category comprised a minimum of 150 businesses. The proportion of firms that were exporters within each of the categories was calculated for the goods-producing and then for the services sectors resulting in the scatter plots shown in Figure 1.

Figure 1: Export Propensity and Firm Size

Source: Canadian SME Exporters (Orser, Spence, Riding and Carrington, 2007)

As Figure 1 illustrates, the scatter supports — for goods-producing firms only — previous results about an S-shaped relationship between export propensity and firm size as measured by the number of employees. For goods producers, a shift in the size-export propensity continuum is observed at about the 25-employee mark. The chart, therefore, confirms previous findings, such as those of Julien et al. (1997), according to which the likelihood of exporting increases substantially once a firm has more than (approximately) 30 employees. For services firms, however, this was not the case and the data do not support the idea of a minimum threshold size.

4.4.2 Estimating Thresholds and the Production Function

To estimate the production function,

where:

output was measured as annual sales revenues. This is not an ideal measure because production theory defines the dependent variable as the volume of production rather than as the value of production; however, sales volume was the closest proxy available from the data. The measure of labour input was the number of full-time equivalent employees and the capital measure was the stock of fixed assets from Part 2 of the FDI data. Management experience was measured as the number of years of experience reported by the primary owner respondent. The estimation was undertaken for four categories of firms as shown in Table 7.

Goods Producers

The production function defined above was estimated for the 549 goods producers for which data were available. The function was then re-estimated separately for (a) the 44 goods exporters in the sample and (b) the 505 goods producers that did not export. The results of each of the three estimations are summarized in Table 8. In all cases, both labour and capital inputs were related to revenues to a statistically and materially significant extent. In neither case was the role of management experience a significant factor of production.

The Chow test was employed to compare the coefficient estimates of the production functions corresponding to the non-exporter and exporter sub-samples of goods producers. The Chow test indicated that the coefficients of exporter firms' production function differ from those of non-exporters at a p-value of 0.013.^{Footnote 17} Goods exporter firms rely somewhat less on labour, and somewhat more on capital, than do non exporters. In neither case was management experience related to output to a statistically significant extent.

Services Firms

The production function above was also estimated for the 1,039 services producers for which data were available (106 services exporters and the 933 services firms that did not export). The results are summarized in Table 9.

As was true for goods producers, labour and capital inputs were related to revenues to a statistically and materially significant extent in all three estimations. Again, use of the Chow test to compare the estimated coefficients finds that the coefficients of exporter firms' production function differ from those of non-exporters, this time at a p-value of less than 0.000. Exporter firms rely somewhat less on labour, and somewhat more of capital, than do non exporters. However, the primary difference between services exporters and non-exporters was in terms of the role of managerial experience. Services exporters rely, to a significantly greater extent, on management experience than do services exporters. It is on this dimension, management experience, that services exporters and non-exporters differ most.

In addition, the data reveal salient differences between goods and services producers. First, for services firms (unlike goods producers) the role of management experience was a significant factor of production. Second, the ratio of labour to capital was considerably higher for services firms than for goods firms. While this might be expected, the difference speaks to differing perceptions of what barr iers to exporting. For goods producers, relatively larger capital stocks are employed relative to services firms; hence, goods producers are likely to be relatively more impacted by perceived capital constraints whereas for services firms, access to skilled labour is more likely to act as a binding constraint.

These are significant findings in terms of its implications for training and public policy because the vast majority of Canadian SMEs are in the services sector. To the extent that they are constrained from exporting due to low levels of management experience a clear training role for educational institutions and governments may be implied through which shortfalls in experience might be remedied through training and education.

4.5 Summary of Study Propositions

Table 10 presents a summary overview of the proposed study propositions and resultant findings.

^{Footnote 15} Financial measures were found to contain a high level of skewness in that a relatively small number of firms reported very high values for revenues and assets. Because of this skewness, statistical testing was not conducted.

^{Footnote 16} The general form of a logistic model is:

log it θ(x)= log [   θ(x)   ] =
    
[ 1-θ(x) ]   

α+β₁x₁+β₂x₂+β₃x₃+ ...

(3)

Where

θ =      e^{(α+β₁x₁+β₂x₂+β₃x₃+ ...)}

1+e^{(α+β₁x₁+β₂x₂+β₃x₃+ ...)}

(4)

Here, the dependent variable was a binary (0, 1) variable corresponding to whether the firm is an exporter (=1) or not (=0). According to the logistic regression framework when the exponent of e in the above equation is large, θ approaches a value of 1 (corresponds to the firm being an exporter). When the exponent of e is small, θ approaches a value of 0, corresponding to non-exporters. The estimates of β_i allow inference about the relative impact of each of the independent variables.

^{Footnote 17} See Chow (1960) "Tests of Equality between Sets of Coefficients in Two Linear Regressions", Econometrica, 28(3): 591–605.

FirstPreviousNextLast