Adoption of New Movie Distribution Services on the Internet

Abstract

Although the Internet features excess demand for media, especially movie downloads, the motion picture industry lacks sustainable business models for this market. An enriched form of the theory of planned behavior can identify drivers of consumer intentions to adopt a legal movie download service. Using a large data set, this study estimates structural equation model parameters. The magnitude of specific influences is subject to unobserved heterogeneity, according to a finite mixture approach with partial least squares estimation. Not only attitude, but also social influence, perceived technical feasibility, innovativeness, compatibility, and past behavior drive adoption decisions.

ACKNOWLEDGMENT

We thank Jan Becker, Christian Ringle, the editors, and three anonymous reviewers for their valuable comments on a previous version of this article.

Notes

¹ Although downloading music and movies became a mass phenomenon in the late 1990s, its beginning dates back to the early days of the Usenet and to initiatives like the Internet Underground Music Archive (Lieb, 1995). We thank an anonymous reviewer for this comment.

² The company wishes to remain anonymous.

³ Oliver Schusser (Apple), Hubert Jakob (T-Online), Christian Hortz (Bertelsmann), Tim van Dyk (Warner Bros.), Sushel Bijganath (RTL), and Andreas Briese (Telepool).

⁴ Exceptions include, for example, CinemaNow and the German Videoload.

⁵ See, for example, Eliashberg, Elberse, and Leenders (2006); Lacy and Bauer (2006); Vogel (2004); and Wasko (1995).

^a Scored on a 5-point Likert scale ranging from 1 (certainly not) to 5 (certainly).

⁶ To ensure the validity of our dependent variable, we correlated intention to adopt with the direct measure of willingness to pay and found a highly significant correlation of 0.41, which indicates reasonable convergent validity.

^a AVE = 0.69; Cronbach's α = 0.54; composite reliability = 0.81.

^b AVE = 0.87; Cronbach's α = 0.85; composite reliability = 0.93.

^c AVE = 0.78; Cronbach's α = 0.72; composite reliability = 0.88.

^d Average variance extracted (AVE) = 0.61; Cronbach's α = 0.80; composite reliability = 0.86.

^e Scored on a 5-point Likert scale ranging from 1 (never) to 5 (often).

^f AVE = 0.63; Cronbach's α = 0.71; composite reliability = 0.84.

⁷ Recent research indicates that in many instances the use of single item measures is appropriate because it does not negatively affect validity (Bergkvist & Rossiter, 2007).

⁸ Further advantages of partial least squares are its less-restrictive assumption regarding sample size and distribution, which are not relevant in our study because the sample size was large enough.

*p < .05.

**p < .01.

⁹ Partial least squares uses unadjusted R ² values.

¹⁰ Following Chin, Marcolin, and Newsted (2003), we computed the effect size reported in Table 3 for each construct as being f ² = (R ² _incl − R ² _excl )/(1 − R ² _incl ).

^a Item loadings associated with a reflective construct.

*p < .05.

**p < .01.

¹¹ To avoid the possibility that multicollinearity causes nonsignificant effects, we applied additional tests for which the variance inflation factors values suggest multicollinearity does not pose a problem.

¹² We thank an anonymous reviewer for pointing out this issue.