Optimal Inter-release Timing for Sequentially Released Products

The pandemic has upended this sequential model with many studios going straight to streaming or simultaneously releasing in theater and streaming services.

The applicability of the model may well go beyond the theater and DVD stages and extend to other stages of a typical Hollywood movie’s sequential release scheme, such as pay-per-view (PPV), video-on-demand (VOD), premium channel premiere, and network TV showing. We focus on the issue of theater-DVD window to simplify the conceptual underpinnings of the econometric approach. Currently, the theatrical and DVD markets combined account for over 90% of the movie-related revenue.

While many executives including the President of Universal Studios Rick Finkelstein perceive that this trend of shortening DVD releases has “gone too far,” others such as Disney have proposed shortening the movie-DVD window to a 4-month standard. At the extreme, Mark Cuban of 29/29 Entertainment has advocated a simultaneous release of movies and DVDs. In fact, Mark Cuban’s studio recently released Steven Soderbergh’s “Bubble” in theaters only 4 days before it became available on DVD, but the movie proved to be a small-scale experiment since it was boycotted by major theater chains.

It took only 5 years for 30 million DVD players to be sold, compared to about 8 years for CD players, and 10 years for PCs to reach the same volume mark.

DVD rentals totaled $5.7 billion, up from $4.5 billion in 2003. Couple that with DVD sales of $15.5 billion, the DVD market over twice as large as the theatrical exhibition market. With DVD penetration spiraling, VHS market has been dwindling: VHS sales dropped 42% to 240.4 million from 2002, while VHS rentals fell 23% to 53.2 million (MPAA 2004). Therefore, the empirical study does not consider the VHS market.

The study does not consider previously viewed DVDs for the following two reasons: first, the sales of previously owned DVDs was approximately $2 billion in 2004; only 7–8% of the $26 billion DVD market. Second, previously viewed DVDs usually contribute revenues to video retailers (or “rentailers”) but not to the studios, so they would have a negligible impact on the studios’ marketing-mix decisions. Nevertheless, some consumers may strategically wait to purchase previously viewed DVDs, and, as a result, the pricing and timing decisions of the new DVD release might have an effect on the incentive to do so. However, modeling such effect requires a different approach that resembles previous models on secondhand markets such as used automobile or textbook markets. And we do not consider catalog DVDs (i.e., DVD release for movies more than 2 years old) for three reasons. First, new-release DVDs account for a large majority of revenues while catalog DVDs represent a small proportion of total pre-recorded DVD sales. Second, since catalog DVDs are released long after their theatrical release dates, the timing decisions are affected by different factors than what is considered in our model; for instance, the DVD of “Assault on Precinct 13” (1976) was released when the remake of the movie was about to open in theaters.

We focus on movies whose box-office gross was above five million dollars because extremely small-budget movies are usually marketed differently (for instance, such movies are targeted at a small niche market and are usually supported by no advertising; they may simply go directly to videos, bypassing the theater channel altogether).

TV is the major channel for DVD advertising, representing 60–70% of the industry spending because of TV’s ability to show DVD trailers.

Such characteristics of movies may include news coverage of the movie and/or tabloid fame of its stars.

Orbach and Einav [34] examine the uniform pricing scheme in the theatrical movie market and argue that this regime is inferior to alternative pricing strategies.

We believe it to be an innocuous assumption; we also estimated a specification without this single-viewing constraint, and the estimation and policy analysis results remain virtually unchanged.

The US Copyright Act of 1976 stipulates that the owner of a legally owned copy of a copyrighted product is entitled to “first use” (commonly known as the First Sale Doctrine), which invokes copyright jurisdiction only upon the first sale of videos so that subsequent usage (such as rental) no longer generates revenue to the copyright holder. This effectively prevents movie studios to discriminate between institutional buyers (i.e., video rental stores) and individual buyers. See [24] for a detailed discussion of its implication on studios’ pricing strategies and the difference between the US market and the E.U. market.

We do not model the case in which the household first rents the video and then buys, or the reverse. We do not think such a simplification severely compromises the validity of the model implications.

Video rental stores typically set a uniform price for all new releases. Therefore, we let \({p}_m^R={p}^R\).

Another way to model such difference is to view the buying utility as a discounted sum of per-period utilities and explicitly specify the discounting patterns [23].

Assuming rational expectations (i.e., the agent’s expectations are objectively correct) is a prevailing practice in dynamic choice economic models. However, such maintained assumptions may be questionable, given that the multiple forms of expectations can all lead to the observed choice behavior (e.g., [38]). It would be ideal if we had data on stated expectations (e.g., how soon consumers expect a particular DVD to be released); however, such questions are not asked in our consumer survey data.

The 2SLS estimates are computed in the first stage by using A = (Z^′Z)⁻¹, then the resulting parameter estimates are used to compute the optimal weighting matrix, \(A={\left({Z}^{\prime}\xi \left({\hat{\theta}}_{2 SLS}\right)\xi {\left({\hat{\theta}}_{2 SLS}\right)}^{\prime }Z\right)}^{-1}\).

Some industry insiders claimed that the trend towards a faster DVD release is caused by an ever-shortening movie leg at the box-office. Our results indicate that the claim is untrue. First, even controlling for the movie leg, the trend variable has a significantly negative coefficient. Second, we also performed a simple regression of the movie leg against a time trend, and the trend variable is not significant, i.e., there is no evidence that movies’ legs have been shortening during our sample period.

Leaving the hardware adoption decision out of the current framework might be problematic if the trend towards a shorter theater-to-DVD window induces consumers to adopt the DVD player earlier than they otherwise would, which subsequently increases the demand for DVD software titles. However, this effect is not identifiable with our current data.