Chapter Listing | Return to 1997 Topical Index



The data analyzed in the present study was compiled in Finland, Sweden, Thailand, and USA in late 1996 and early 1997. The respective universities are Helsinki University of Technology in Finland, Linköping University in Sweden, the University of Colorado in USA, and the Asian Institute of Technology in Thailand. The combined sample size is 1956 university students. The students are mostly students of technology. The respondents were chosen randomly from the general student population. The students were mailed or interviewed with a four-page questionnaire. The questionnaire was first designed in English and then translated to local language if other than English. At the Asian Institute of Technology (AIT), the English language version of the questionnaire was used, as the teaching language at AIT is English. The response rates varied from 24 % to 53 % in mailed surveys (Helsinki, Colorado) and from 75 % to 90 % in interview surveys (Linköping, Bangkok).

The basic statistics of the empirical sample are shown in tables 1 and 2. Although the sub-samples were designed to be as similar as possible, some differences remain. The ages of the respondents varied slightly. The Asian and American respondents were oldest, with a median age of 27 - 28 years. The Swedish sample was considerably younger, with a median age of 21 years. The Finnish respondents were, on the average, 25 years old. The older respondents were more likely to be married, have children, be graduate students instead of undergraduates, and they had more work experience.

The employment situation of the respondents at the time of the study varied as well. Of the young Swedes, only 20 % were working either part-time or full-time. Also in Asia, the fraction of respondents employed at the time of the survey the questionnaire was low, 30 %. In Finland, 56% of the respondents said they were working either full time or part time, whereas in the Colorado sample, the figure was 82 %.

The career preferences were to some degree quite similar in the different databases, in that civil servant careers and academic careers were the least favored in all sub-samples. The two favorite career choices were corporate career and entrepreneurial career. In Sweden, and especially in Finland, the corporate career was the clear favorite. In Asia and the U.S., the preference for the corporate and entrepreneurial careers was equal.

Testing the Model

Path analysis was applied on testing the model. First, the influences of various background variables on entrepreneurial conviction were tested. Second, the influences of the same variables, complemented with entrepreneurial conviction, were tested on entrepreneurial intent. In the first phase of the analysis, the combined sample of American, Scandinavian, and Asian students was used. The robustness of the resulting models was then checked by running the same analyses for the different country databases.

The results of multiple regression analysis on entrepreneurial conviction (using backward elimination) for the combined database are shown in Table 3. As table 3 shows, general attitudes have a high moderating influence on entrepreneurial conviction. In particular, the need for achievement and autonomy emerge as influential attitudinal moderators of conviction. As anticipated in the base model, image-payoff is also indicated a high Beta. Also the perceived support of the university environment emerges as a moderating influence. A number of background variables are also included in the model, the most important of these being previous work experience in SMEs, gender (higher values indicated for males), and vicarious experience. All in all, the model explains approximately 40 % of the variance in conviction.

The influences on entrepreneurial intent for the combined sample are shown in table 4. In this analysis, the expressed intent to start working for one’s own firm on a full time basis within one year is used as the dependent variable. The strength of this intent was expressed with a 4-point Likert scale. The model in table 4 is in general agreement with the base model. Conviction stands out as clearly the most important influence, with a Beta of 0.347. The influence of age is also fairly high, as is the influence of previous work experience in SMEs. As the number of years studied also appears as a moderating influence, we are tempted to interpret these age related influences as situational. The likelihood of imminent graduation tends to correlate strongly with age and with the number of years studied.

The model in table 4 supports the expectation that the influence of attitudinal variables is moderated by conviction. However, the influence of university environment appears strange. A negative, albeit fairly weak, influence is indicated for the support of university environment on entrepreneurial intent. This may be due to the operationalization of entrepreneurial intent as the perceived likelihood of working full time for one’s own firm within one year from the point of time when the survey was carried out. Also note that the explanatory power of the model in table 4 is fairly low, only 22 %. This is a lower explanatory power than the ones established in previous studies that have used a looser operationalization of entrepreneurial intent. Krueger (1993) established a R square of 0.543 for his full model. Davidsson (1995) established a R square of 0.51 for his all variables model. Reitan (1996) achieved a R square of 0.63 for his all variables model. However, when a more strict short term intention is used as a dependent variable, the explanatory power of the model of the present study reaches the same class of magnitude. Davidsson established an explanatory power of 0.32 for his base model, when one-year intentions were used as a dependent variable. Reitan established an explanatory power of 0.30 for his all variables model when two-year intentions were used as a dependent variable.

Robustness of the model

The robustness of the models was checked by running the same analyses for each country sample. Table 5 shows the resulting models for conviction. Table 6 shows the resulting models for intent. The robustness checks in tables 5 and 6 confirm the main findings in a surprisingly uniform way. Only Sweden stands out as anomalous. This is due to technical issues in data collection, which render the Swedish sample effectively incomparable with the other samples. In the following discussion, the Swedish data should be ignored.

The need for achievement and autonomy emerge as the most important influences in both Finnish, US, and Asian samples. In addition, all country samples confirm the importance of supporting university environment as an important influence on conviction. The Finnish and US samples further confirm the importance of Image-Payoff as an important influence. This variable does not emerge as an influence in the Asian data. Surprisingly, gender emerges as an influence in the Finnish and US data, but not in the Asian data. If anything, an opposite influence could have been expected here. This influence starts to make more sense when the check runs in table 6 are reviewed, however.

For short term entrepreneurial intent, the importance of conviction as a moderating influence on intent is uniformly confirmed. The magnitude of this influence is approximately similar for the Finnish and US samples (Betas 0.31 and 0.34, respectively), and even stronger for the Asian sample (Beta 0.47). Other variables are indicated variable influences in the country samples. A point of interest is found in the Asian sample, where a fairly strong influence of gender is detected, with females signaling stronger intent. Perhaps this finding signals the attempt of females to avoid the glass ceiling of Asian hierarchical organizations. This finding could also explain why the Asian females did not signal lower levels of entrepreneurial conviction, as did their counterparts in the Finnish and US samples.

Summary and DISCUSSION

The findings of the path analysis are summarized in figure 2. The analysis confirms many previous findings in the literature. The findings provide support for the usability of the process approach to analyzing entrepreneurial behavior. So far, the tests of the process approach have been limited to samples collected in homogenous cultural environment. Our study contributes to this literature by demonstrating the robustness of the intent approach in different cultural environments. The robustness checks using different country samples provide indicate remarkable uniformity in the country samples, considering that the samples have been compiled in highly diverse cultural environments. The central expectations are uniformly confirmed.



Illustration of the final model, relationships grouped

The model constructed in the study has also policy implications. There are numerous policy initiatives, such as business incubator programs, that support the emergence of new, technology-based firms from universities. Quite often, such programs try to influence behavior only, not intent and other cognitive factors that influence behavior. The findings of the study provide pointers for expanding the scope of policy initiatives. Our data shows that entrepreneurial conviction among university students is influenced by the image of entrepreneurship as a career alternative, as well as the encouragement and support received from the university environment. In our study, as well as in earlier studies, conviction emerges as the most important influence on intent. As entrepreneurial intent is a (supposedly) good predictor of entrepreneurial behavior, our findings suggest that spin-off support programs could also employ indirect approaches to pursuing their goals. Entrepreneurial conviction among university students could be influenced by fostering encouraging attitude toward SMEs in universities. Such action could entail, for example, establishing and signaling clear IPR policies at the university, using positive role models in teaching, and fostering a positive image of entrepreneurship as a career alternative. Our findings suggest that such measures should have an indirect positive influence on entrepreneurial behavior.

Note that the intent measure used in the present study has been a fairly strict one. Our operationalization comes close to the definition of nascent entrepreneur. This approach has both merits and weaknesses. A merit of this operationalization is that its predictive power is probably stronger than those of the more loose operationalizations. A weakness of the selected approach is that the occurrence of strong intent of this kind is relatively rare. The weaknesses of the use of multiple regression techniques in analyzing strongly biased distributions have been demonstrated by, e g, Reynolds (1995), who suggests the use of automatic interaction detection techniques for the analysis of such situations. We will do this in the future versions of this paper. We also plan to extend the analysis by running the analyses using alternative measures of intent as dependent variable.

So far, the research on entrepreneurial intent has, by necessity, had to assume that intentions predict entrepreneurial behavior. Actual data to confirm this assumption remains scarce, however. There is plenty of evidence emphasizing the importance of situational contingencies on entrepreneurial behavior. Future studies should strive to build on previous intent surveys to check, to what extent entrepreneurial intent is followed through. Our intent is to follow up on our database to check the predictive power of the intent measures used in this study.

Top of page | Chapter Listing | Return to 1997 Topical Index

© 1997 Babson College All Rights Reserved
Last Updated 03/02/98