The Use of the Internet in Distributing Packaged Software
To reflect common practice in the software industry and extend transaction cost theory, this research developed building on previous research and empirically tested a model based on to identify conditions in which software vendors are likely to sell and distribute their packaged products directly to end-users through the Internet. How software firms distribute their products over the Internet is an important issue because software is a digital product, and the potential for the Internet to transform the distribution channel is considerable. Extant literature shows that Canadian software firms frequently choose direct instead of market channels. However, none of the existing studies focuses specifically on packaged software, or on the Internet as a distribution channel. Further, recent research on what products are suitable for distribution through the Internet does not address the case of packaged software. <ul> <li>Knowledge-based asset specificity, human asset specificity, and physical asset specificity are positively associated with the likelihood of using the Internet to distribute packaged software (H<sub>1</sub>, H<sub>2</sub>, and H<sub>3</sub>). </li> <li>The likelihood of using the Internet in delivering products has a positive relationship with the volatility of packaged software, its clients, and markets (H<sub>4</sub>); whereas this likelihood has a negative relationship with diversity (H<sub>5</sub>). </li> <li>Channel growth is positively associated with the online distribution of packaged software (H<sub>6</sub>); Channel volume is negatively associated with the likelihood that packaged software developers use the Internet to deliver products (H<sub>7</sub>). </li> <li>The rate of growth in gross sales has a positive relationship with the likelihood of online distribution by packaged software firms (H<sub>8</sub>); while the gross sales of a firm negatively are associated with this likelihood (H<sub>9</sub>). </li> <li>The use of the Internet in the distribution of packaged software is positively associated with the United States market and negatively associated with other national markets (H<sub>10</sub>). </li> </ul> The data to test these hypotheses were collected from Canadian software developers by a web-based survey. The information includes the distribution channels for their best selling product in its largest market, and Likert scales that measure forms of asset specificity, market uncertainty, and channel volume. The hypotheses are tested using logistic regression. The results provide support for hypotheses H<sub>5</sub>, H<sub>6</sub>, and H<sub>9</sub> whiles hypotheses H<sub>1</sub>, H<sub>2</sub>, H<sub>4</sub>, H<sub>7</sub>, H<sub>8</sub>, and H<sub>10</sub> are not supported. The result for H<sub>3</sub> is statistically significant, but the direction of the relationship is the opposite of the expectation. <br /><br /> The results of this study have implications both for theory and managerial practice. This research contributes to the literature a test of the ability of transaction cost analysis to explain the use of the Internet in distributing software. It also provides managers with reliable insights into some of the circumstances where packaged software developers may use the Internet to deliver their products. However, further research is required to verify the generalizability of the findings of this study.