WEB APPENDIX Managing Innovation Sequences Over Iterated Offerings: Developing and Testing a Relative Innovation, Comfort, and Stimulation Framework of Consumer Responses Timothy B. Heath Subimal Chatterjee Suman Basuroy Thorsten Hennig-Thurau Bruno Kocher
Innovation Valence WEB APPENDIX A The Independence of Innovation Magnitude and Innovation Valence Followed by Each Study s Focus Innovation Magnitude Minor Major Weaker (Less Liked) Stronger (More Liked) Weaker Minor Stronger Minor Weaker Major Stronger Major Prototypic Approach Study 1 Three studies compare (1) some version of an innovation sequence starting with a weaker-minor innovation followed by a stronger-major innovation and (2) some version of an innovation sequence starting with a stronger-major innovation followed by a weaker-minor innovation. Study 1 experimentally varies innovation valence and sequence within a gaming context with innovations that are minor-to-moderate in magnitude. Innovation Sequences: Weaker-Stronger vs. Stronger-Weaker Study 2 Study 3 Study 2 experimentally varies innovation valence and magnitude together as well as innovation sequence within three product contexts (Weaker = Minor, Stronger = Major). Innovation Sequences: Weaker/Minor-then-Stronger/Major vs. Stronger/Major-then-Weaker/Minor Study 3 models the effects of innovation magnitudes and sequences on box office sales of film sequels whose serial positions (order in the film franchise) differ. Innovation Sequences: Minor-Major vs. Major-Minor constitute the focal sequences, though all sequences are possible depending on how the franchises evolved. Innovation magnitude is operationalized as the number of innovations made on seven key film dimensions.
WEB APPENDIX B Study 1 Example RPS Outcome Sequences: Banter-then-Winning-Streak Sequence in the Cumulative Condition Cond Innov. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Summary Period 1 Baseline W L T T W L T T W L L T W T Ahead: 5 Order-A 1-0-0 1-0-1 1-1-1 1-2-1 2-2-1 2-2-2 2-3-2 2-4-2 3-4-2 3-4-3 3-4-4 3-5-4 4-5-4 4-6-4 Behind: 5 15 16 17 18 19 20 21 22 23 24 25 26 27 28 Tied: 18 L W T T L W T T L W W T L T 4-6-5 5-6-5 5-7-5 5-8-5 5-8-6 6-8-6 6-9-6 6-10-6 6-10-7 7-10-7 8-10-7 8-11-7 8-11-8 8-12-8 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Period 2 Banter L W T W T L L W L T W W T L Ahead: 7 Order-B1 0-0-1 1-0-1 1-1-1 2-1-1 2-2-1 2-2-2 2-2-3 3-2-3 3-2-4 3-3-4 4-3-4 5-3-4 5-4-4 5-4-5 Behind: 7 15 16 17 18 19 20 21 22 23 24 25 26 27 28 Tied: 14 T L W W T L L W L W T W L T 5-5-5 5-5-6 6-5-6 7-5-6 7-6-6 7-6-7 7-6-8 8-6-8 8-6-9 9-6-9 9-7-9 10-7-9 10-7-10 10-8-10 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Period 3 Banter W L T L T W W L W T L L T W Ahead: 9 Behind: 8 Order-B2 Streak 1-0-0 1-0-1 1-1-1 1-1-2 1-2-2 2-2-2 4-2-2 4-2-3 5-2-3 5-3-3 5-3-4 5-3-6 5-4-6 6-4-6 Tied: 11 15 16 17 18 19 20 21 22 23 24 25 26 27 28 T W L L T W W L W L T L W T 6-5-6 7-5-6 7-5-7 7-5-9 7-6-9 8-6-9 10-6-9 10-6-10 11-6-10 11-6-11 11-7-11 11-7-12 12-7-12 12-8-12 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Period 4 Banter T L W T W W L L T L T W T L Ahead: 6 Order-C Streak 0-1-0 0-1-1 1-1-1 1-2-1 2-2-1 4-2-1 4-2-2 4-2-4 4-3-4 4-3-5 4-4-5 5-4-5 5-5-5 5-5-6 Behind: 6 History 15 16 17 18 19 20 21 22 23 24 25 26 27 28 Tied: 16 W T W T L T L T W T W L T T 6-5-6 6-6-6 7-6-6 7-7-6 7-7-7 7-8-7 7-8-8 7-9-8 8-9-8 8-10-8 9-10-8 9-10-9 9-11-9 9-12-9 Notes: Innov = Innovation. The numbers 1-28 designate the 28 plays within each of the four periods. W = win, L = loss, T = Tie. Green Letter Color = participant wins that play, Black Letter Color = participant ties that play, Red Letter Color = participant is loses that play. Green Numbers Color = participant is winning overall, Black Numbers Color = participant is tied overall, Red Numbers Color = participant is losing overall. Numbers in each cell indicate the score after that round has finished in order of wins, ties, losses.
4 WEB APPENDIX C TABLE C1 Study 1: Entertainment Levels and Changes by Condition and Period Introduction-Weaker-Stronger Sequence Introduction-Stronger-Weaker Sequence Cumulativeness Value P2 P3 P4 P23- P3- P4- P2 P3 P4 P23- P3- P4- Non-Cumulative Entertainment 3.65 3.37 4.00 3.21 0.03 0.35-0.44 4.09 4.18 3.32 3.44-0.34-0.76-0.64 SD 1.47 1.52 1.62 1.57 0.78 1.10 1.00 1.71 1.67 1.79 1.75 0.72 0.96 1.29 n 65 65 65 65 65 65 65 75 75 75 75 75 75 75 Cumulative Entertainment 4.04 3.71 4.17 4.01-0.10 0.13-0.03 3.88 4.15 3.71 3.65 0.05-0.16-0.23 SD 1.47 1.56 1.52 1.59 0.91 1.12 1.34 1.47 1.60 1.78 1.86 0.91 1.14 1.26 n 77 77 77 77 77 77 77 73 73 73 73 73 73 73 Notes: = Period 1, P2 = Period 2, P3 = Period 3, P4 = Period 4, P23 = the change in entertainment when going from Period 1 to Periods 2 and 3 (P3 and P4 differences shown as well), SD = standard deviation, and n = cell size. The RPS game tends to produce more mechanical-turk cell-size variance than we see with other/shorter experiments, but the cells sizes and Period-1 entertainment levels are not correlated with effects (e.g., in a mechanical turk replication not reported here, the non-cumulative weaker-stronger condition was the second largest and yet produced both the best Period-1 performance and best performance in Periods 2 and 3 relative to Period 1). TABLE C2 Study 2: Descriptive Statistics by Innovation Sequence and Period Product and Values Introduction-Weaker- Weaker Sequence Introduction-Weaker- Stronger Sequence Introduction-Stronger- Weaker Sequence Overall P2 P3 P4 P23 P2 P3 P4 P23 P2 P3 P4 P23 Product Interest 3.76 2.81 2.42 3.28 4.21 5.34 3.07 4.77 5.22 2.95 1.73 4.08 SD 1.28 1.73 1.79 1.32 1.51 1.41 1.43 1.25 1.19 1.24 1.15 1.03 n 30 30 30 30 30 30 30 30 30 30 30 30 Film P2 P3 P4 P23 P2 P3 P4 P23 P2 P3 P4 P23 Product Interest 3.64 1.80 1.22 2.72 4.65 5.19 3.05 4.92 5.48 3.17 1.59 4.32 SD 1.80 1.54 1.29 1.51 1.26 1.05 1.40 0.94 0.81 1.00 1.30 0.83 n 6 6 6 6 11 11 11 11 13 13 13 13 Music P2 P3 P4 P23 P2 P3 P4 P23 P2 P3 P4 P23 Product Interest 4.19 2.85 2.44 3.52 4.10 4.78 3.06 4.44 4.74 3.46 1.82 4.10 SD 1.20 2.01 1.95 1.48 1.69 1.40 1.47 1.39 1.52 1.03 1.04 1.08 n 12 12 12 12 11 11 11 11 7 7 7 7 Tablet P2 P3 P4 P23 P2 P3 P4 P23 P2 P3 P4 P23 Product Interest 3.39 3.26 3.01 3.32 3.76 6.29 3.10 5.03 5.22 2.30 1.85 3.76 SD 1.02 1.41 1.67 1.08 1.59 1.50 1.60 1.47 1.36 1.46 1.12 1.23 n 12 12 12 12 8 8 8 8 10 10 10 10 Notes: P2 = Period 2, P3 = Period 3, P23 = Periods 2 and 3 combined, P4 = Period 4, SD = standard deviation, n = cell size.
5 WEB APPENDIX D Study 2: Example Experimental Conditions Film Example: Introduction-Weaker-Stronger Innovation Sequence Now imagine that 18 months have passed. A sequel to the movie noted on the prior page will be in theaters soon. It is a quality film and has the following features relative to the original film: Story Characters Actors Context Similar and Moderately Mostly the Same The Same Similar Predictable (measures) And now imagine that another 18 months have passed. A third film in the series will be in theaters soon. It is a quality film and has the following features relative to the prior (second) film: Story Characters Actors Context Three New Characters (Two Family The Same Plus Members) that Re-cast the Stories and New Ones for Change Interpersonal Dynamics New Characters Different Stories that are Not Predictable Music Example: Introduction-Stronger-Weaker Innovation Sequence Partially the Same, Partially Different Now imagine that two years have passed. The band is about to release its second album. It is a quality album and has the following features relative to the band s first album: General Sound (Instruments, Style) Melodies and Rhythms Lyrics Different Instruments and Electronically Created Some Similar, Others Quite Broader Topics Sounds; Somewhat Different Style Different (measures) And now imagine that another two years have passed. The band is about to release its third album. It is a quality album and has the following features relative to the band s prior (second) album: General Sound (Instruments, Style) Melodies and Rhythms Lyrics The Same Similar Similar Computer Tablet Stimuli: Clamshell Design Intro-Weaker-Weaker Period-2 Iteration Intro-Weaker-Weaker Period-3 Iteration Size & Size & Exterior Shape Functions Exterior Shape Functions The Same but Now Comes in Multiple Colors The Same Minor Improvements (new apps available, more memory, etc.) The Same but Now Available in Stainless Steel (natural silver or black) The Same Minor Improvements (new apps available, more memory, etc.) Intro-Weaker-Weaker Period-4 Iteration Stronger Innovation Size & Size & Exterior Shape Functions Exterior Shape The Same The Minor Cutting Edge Smartphone Built-in to Same but but Now Same Improvements the Tablet. Phone Slides Out and Phone Comes with (new apps Can be Used as Primary Cell Phone Slides into an Over-the- available, more (all major phone carriers accepted; a Docking Shoulder memory, etc.) tablet automatically backs up phone Slot Carrying data; widespread internet access for Case both; computer can charge phone battery) Functions New Apps, More Memory, New Software Integrates Tablet and Phone
6 WEB APPENDIX E Study 3: Endogeneity, Estimation, Diagnostics, and Robustness Checks Endogeneity. Several potential endogeneity effects could bias ordinary least squares in this context. First, endogeneity between revenues and screens is possible (Elberse and Eliashberg 2003). We use two instruments for the number of screens of the current film: (1) number of screens available for the previous film in the franchise, and (2) the total number of screens available for the top 10 movies in the week the focal film is released. The key idea for these instruments is that the number of screens dedicated to the previous film in the franchise, and number of screens devoted to the top ten films in the same week, should affect the number of screens allocated to the focal film, but should not affect the focal film s revenues. The second potential source of endogeneity is the innovation variable itself. Innovation is endogenous because omitted variables can prompt studios to make changes and affect the revenues as well. We use several instruments for innovation: lagged (previous sequel) ROI (and square of lagged return), and lagged innovation. The prior sequel s ROI and innovations should affect the innovations made in the focal film, but they are unrelated to the focal film s revenues. The third potential source of endogeneity is the production budget (CPI adjusted), as weaker films (in terms of box office capabilities as perceived by the studios) will be assigned smaller budgets by studios. As an instrument for budget, we use the number of films released by the studio in the current year. The logic is that, at the studio level, the budget is a resource allocation issue (Vogel 2001), and given that resources are limited, the budget allocated to any specific film is likely to depend on the studio s portfolio of films. However, the number of films in the studio s portfolio is unrelated to the current film s revenues (it affects revenues only through the budget allocated to it). Finally, the advertising budget may also be endogenous for two reasons. First, firms may have spent more on advertising for later sequels, though the low negative correlation between advertising budget and serial position (r =.24) contradicts this possibility. Second, firms may have allocated larger advertising budgets to major (versus minor) innovations, but again the low correlation between advertising budget and innovation magnitude (r =.03) suggests otherwise. Consistent with these results, an additional explicit test for advertising budget endogeneity proved negative. Specifically, we calculated the GMM C statistic (i.e., the gmm equivalent to the Wu-Hausman test for endogeneity) using all exogenous variables from the demand equation plus the prior film s advertising budget as instruments for the current film s advertising budget (the prior film s advertising budget is a good predictor of the current film s advertising budget yet uncorrelated with the current film s financial success). The GMM C statistic fails to reject the null hypothesis of advertising budget exogeneity (GMM C statistic chi2(1) = 1.80; p =.18). These three results are consistent with Clement, Wu, and Fischer s (2014) conclusion that advertising can be safely treated as exogenous in film contexts. Estimation. Although the instrumental techniques allow addressing the correlation between the predictors and the error term, there is still the issue of correlated error terms for each movie across the franchise. Given that movies within each franchise (or cluster) may be correlated because of serial correlation or random effects, we use the GMM (Generalized
7 Method of Moments) estimation technique (for details, see Greene, 2012). The different steps are (1) estimate Equation (1), or the basic regression model using standard instrumental variables methods, (2) use the residuals from the first step to obtain the optimal GMM weighting matrix, and (3) allow for heteroskedasticity and correlation between error terms for movie-franchise combinations for every movie. Diagnostics. We checked for various regression diagnostics including (a) heteroskedasticity (the Breusch-Pagan test rejects homoscedasticity (χχ 2 (1) = 56.40, p <.00) 2 leading us to use the GMM estimation technique) (b) endogeneity effects (both the Durbin (χχ (3) = 21.49, p < 0.00) and the Wu-Hausman (χχ 2 (3) = 7.57, p < 0.00) tests reject the null or exogeneity of the regressors), and (c) the instruments (the Stock and Yogo s F statistics are over 10 in the first step (instrumental variables regressions), which demonstrate that the instruments chosen are relevant (Wooldridge 2009), and the Hansen-J statistic is insignificant (p >.10; Wooldridge 2003) showing that the instruments are also exogenous). Robustness Checks. We performed a number of robustness checks on our results. First, a key dimension of the innovation variable is the sequel s star cast. The analysis in Appendix A codes star change based on whether the top star remained the same or differed from the predecessor. The results of our analysis did not change when we based it upon the either one of the top two stars (in billing order, as listed in IMDb database). Second, one might suspect that studios may have used more popular stars when making changes to later sequels (i.e., change in stars is confounded with the popularity of the stars). To allay this concern, we examined the popularity of the top star (in billing order) at the time of the release of the sequels as provided by the IMDb Starmeter ratings (the start year is 1998; for movies released prior to 1998, we defaulted to this year). We found no significant differences in star-power between earlier and later sequels (e.g., the average star-ranking for the first two sequels in our data set was 109 and the average star ranking for the third and fourth sequels in our data set was 134; we should note that higher number denote lower star power). Third, another key dimension of the innovation variable is the sequel s genres. The analysis in Appendix A codes genre change based on the Netflix reports. Since IMDb codes movie genres somewhat differently than Netflix, we also incorporated the genres according to the IMDb classification. Our results remained unchanged. Fourth, we checked whether a strategy of enrichment of the genres by later sequels (beyond a simple change picked up by the genre change dummy) might have affected the results. For example, studios may have added more genres when making changes to the later sequels (i.e., change is confounded with enrichment). However, our analyses show that the average number of genres remained unchanged across the life cycle of the franchise (approximately 4.3 genres on the average for both earlier and later sequels).
8 WEB APPENDIX F Study 3: Descriptive Statistics Variable Observations Mean Standard Deviation Minimum Maximum Budget ($Million) 329 59.24 58.53 0.02 316 Advertising ($Million) 332 22.90 16.40 0.04 63.70 Screens 341 2242.00 1088.00 9.00 4468.00 Audience Rating (1-5) 334 3.52 0.35 2.60 4.40 Professional Rating (1-5) 337 3.24 0.42 2.20 4.50 Season (0-1) 341 0.65 0.14 0.38 1.0 Competition 318 6.04 3.65 1.00 19.0 Time Interval (Days) 249 1374.00 1344.00 105.00 9209.00
9 Variables Total Revenue Total Revenue 1.00 ROI ROI.39 1.00 Screens Screen.16.33 1.00 Innovation Innovation -.11 -.03 -.13 1.00 Budget Budget.49 -.27.31 -.17 1.00 Advertising.07 -.09.34.03.50 1.00 Serial Position Professional Rating Audience Rating Advertising WEB APPENDIX G Study 3: Correlations Serial Position -.08 -.03 -.06.16.16 -.24 1.00 Professional Rating.06.17.09 -.12.08 -.03 -.02 1.00.23.01 -.01 -.15.03.21 -.01.83 1.00 Competition -.04.02 -.04 -.17 -.12 -.16.08 -.25 -.08 1.00 Season.13.02.03 -.09.05.15 -.08.03.12 -.14 1.00 R-Rating -.15.09 -.05.01 -.12 -.18.04 -.04 -.07 -.07 -.20 1.00 GPG-Rating.19.06 -.07 -.13 -.10.00.05.18 -.01 -.07 -.16 -.45 1.00 Action- Adventure Release Year Time interval -.01 -.19 -.07.07.34.08 -.01.21.22 -.13.23.01.15 1.00 -.15 -.31.52.11.16.39.09.20.28.69 -.13 -.32 -.27.05 1.00.08 -.11 -.02.15.16.06.17 -.03 -.08 -.07.12.14 -.07.07.11 1.00 Audience Rating Competition Season R-Rating GPG Rating Action Adventure Release Year Time Interval