Mark J. Ferrari Andrew Rudd Executive of Research Managing Accomplice mferrari@procinea arudd@procinea - PowerPoint PPT Presentation

slide1 l.
Skip this Video
Loading SlideShow in 5 Seconds..
Mark J. Ferrari Andrew Rudd Executive of Research Managing Accomplice mferrari@procinea arudd@procinea PowerPoint Presentation
Mark J. Ferrari Andrew Rudd Executive of Research Managing Accomplice mferrari@procinea arudd@procinea

play fullscreen
1 / 36
Download
Download Presentation

Mark J. Ferrari Andrew Rudd Executive of Research Managing Accomplice mferrari@procinea arudd@procinea

Presentation Transcript

  1. Investing in Movies Presentation to the Institute for Quantitative Research in Finance Procinea Management, LLC April 4, 2006 Mark J. Ferrari Andrew Rudd Director of Research Managing Partner mferrari@procinea.comarudd@procinea.com Copyright © 2006

  2. Motivation Introduction • Movies are an Alternative Asset Class • Beneficiary of significant co-financing over the years, mostly tax-driven • Total return to the asset class not particularly exciting • Research project to construct an active quantitative strategy to deliver attractive returns to investors • Part of a broader initiative to analyze investment potential of Artistic and Intellectual Property

  3. Identifying an Investment Strategy Introduction • Two major issues • Define appropriate contract to align interests between investors and studios • Identify investment strategy • Interestingly, it appears relatively difficult to align interests • Investors wary because of unsuccessful attempts by previous investors • Participation accounting strongly favors the studio • Investors have historically borne studio error, bad judgment and profligate spending • Motion picture performance is wholly unpredictable • Studios saddle investors with “losers” • Key steps in strategy research • Define Universe • Rank assets according to “valuation” model • Construct portfolio • Observe performance

  4. Investing in Movies: Agenda Introduction • Movie Industry • Prior Academic Research • Procinea Model Estimation • Strategy Results • Extensions

  5. Industry Players Movie Industry • Focus on movies distributed by the major studios and their subsidiaries • Ensure the movies are contracted for world-wide distribution • Very different from the “Indies” • Lab for behaviorists • Creative vs. business skills • Better to fail spectacularly than succeed modestly? • An isolated success becomes conventional wisdom • “It is a miracle that any movie is ever made…” • It is not just the box office

  6. Worldwide Filmed Entertainment Revenue Movie Industry

  7. Studio Capital Requirements Movie Industry • Major studios (Fox, Disney, Paramount, Sony, Universal and Warner Brothers, and their subsidiary brands) finance 100-125 titles per year • Average production and distribution cost of each title exceeds $100 million, creating annual funding needs of $10 - $12 billion • The studios are voracious users of capital. There is a long history of studios using co-financing partners • Procinea estimates a funding gap of $5 billion or more per year—and this gap is liable to increase as the number of films and costs per film rise

  8. Average Cost of Major Studio Feature Films Movie Industry

  9. Production Cost and Revenue Movie Industry • Average production cost of $60-70m • Production cost incurred over 12-18 months • Prints and Advertising (“P&A”) can be as much or more than production cost • Revenue earned relatively quickly after release, but with a long tail • 60% of revenue in first year of release • Almost 90% by end of second year • We focus on “first cycle” revenues, ignoring library value

  10. Film Life Cycle Movie Industry 100% Domestic & Foreign TV % of Revenue Domestic Pay TV 75% Foreign Home Video / DVD 50% Domestic Home Video / DVD Foreign Box Office 25% Domestic Box Office -18 -12 -6 0 6 12 18 84 Months % of Expenditures Pre-Production & Production 25% 50% Post Production 75% Release Campaign (P&A) 100%

  11. Adverse Selection? Movie Industry • Studios demand increasing financing to fill distribution windows so require outside investors • Investors could wait until the movie is finished before investing, but this subjects them to adverse selection • Assumes that studios have superior information • The stories of Starman and Titanic • Alternatively, investors could partner with the studio at “greenlight” and bear risk pari passu with the studios

  12. Movies Released from 1997 to 2004 Movie Industry • Broad universe 1627 movies • Movies released theatrically in U.S. or Canada • Minimum production cost $2m • Exclude unrated and foreign productions • Estimation universe 836 movies • Minimum production cost $15m, maximum $125m • Exclude sequels, animation, documentary, and NC-17 • Target universe 588 movies • Financed (or co-financed) and distributed by a major studio

  13. Distribution of Target Universe ROI Movie Industry

  14. WOW! Movie Industry • Why so many losers? • Approximately 60% of movies fail to cover costs • Suggests that studios cannot predict the outcome of a movie • Danger of adverse selection possibly not as great as previously thought • Also suggests an active investment strategy • Forecast (net) revenues for each movie • Only invest in those movies where forecast revenues are “large enough”

  15. Implications for Modeling Movie Industry • Limited universe of movies • Revenue is clearly a non-linear function of movie attributes • Interactions between the attributes likely to be very important • Many interesting movie attributes are not publicly available (e.g., actor compensation) • Not an optimistic sign that the studios cannot predict success • Academic literature generally negative

  16. Prior Academic Research Prior Academic Research • Substantial academic literature • Recent works that analyze financial characteristics • Ravid (1999), Postrel (2000), Vogel (2001), De Vany (2004) • Findings include • Highly skewed distribution of returns • Large budgets, movie stars no guarantee of success • Little evidence that movie attributes affect performance • Attributes studied include • Budget, stars, sequels, genre, ratings, screens, box office life, year of release • Connection to literature on project finance (Berk et al.)

  17. Motivating Quotes Prior Academic Research • “…most major-distributed films do no better than financially break-even,” Vogel (2001), p.97 • “Most movies are unprofitable. Large budgets and movie stars do not guarantee success. Even a sequel to a successful movie may flop,” De Vany (2004), p.82 • “…forecasting revenue is futile…,” De Vany (2004), p.90 • “The financial performance of a movie is unpredictable because each one is unique…” Vogel (2001), p.97 • “There are no formulas for success in Hollywood,” De Vany (2004), p.98 • “Most stars do not really make a difference,” Postrel (2000) • “Nobody knows anything,” Goldman (1983)

  18. Revenue Modeling Model Estimation • Total revenue is driven by audience appeal • Why not model return? • Fractional share of investment in each project is fixed • Advertising effect in denominator mitigates outliers • Guiding principles for sparse data • Sensible • Simple • Stable

  19. Data Sources Model Estimation • Unfortunately no single data source is complete • Collect data from standard industry data sources • Augmented with extractions from on-line entertainment media, media research reports, etc. • Define and collect movie attributes not provided by vendor and industry sources • E.g., cast billed order, story elements, etc • Data cross-referenced and cross-validated to achieve a “Compustat-like” database • Includes more than 7,800 films, up to 70 data points/movie • Many interesting items still confidential to the studios • E.g., star compensation

  20. Production Cost Is an Important Factor Model Estimation • Power law with exponent greater than one • In-sample explanatory power is a strong function of range of cost • Lower-budget movies are riskier • Prior to video release, recent movies appear to underperform Data: March 2005

  21. Locally Weighted Regression Model Estimation • Procinea’s proprietary hit ratio D quantifies past financial performance of director • Correlation of cost C and hit ratio is 0.34 • Both factors and their interaction are significant predictors of revenueRi of movie i according to OLS • Naturally handle interaction and heteroskedasticity Tricubic weight Euclidean distance

  22. Director’s Track Record Matters Model Estimation • Better director increases log(Revenue) for any budget • A little skill really helps a small project • A little cash really helps a struggling director • Excellent directors cannot outperform if cash-constrained Data: September 2005

  23. Residual Revenues Are Semi-Lognormal Model Estimation • Fit cost-dependent risk σ(C)to squared residuals after accounting for production cost and director • Distribution of standardized residual ηi/σ(C)

  24. Revenue Depends on Season and Rating Model Estimation • Mean residual ηiand fraction of movies in each cell • Typical standard error of mean residual = 0.05 • Season and rating effects are significant at the 95% level • Interaction of season and rating is not significant

  25. Consistent Abnormal Revenue of Genres Model Estimation • Primary genre from Nielsen • Genres combined until none contains fewer than 30 movies • Mean residual in each genre computed for two sub-periods • Analysis of variance shows significant explanatory power in each sub-period • Correlation shows significant persistence

  26. Quality Is Rewarded Model Estimation • Metascore® between 0 and 100 from metacritic.com • 763 movies from estimation universe, 1995-2004 • Perfect foresight test, not model component • Like a valuation ratio, performs better when genre neutral • Action, Comedy, Drama, Horror, Romance, Sci-Fi, Thriller, and Other • Two standard deviations is an 89% increase in revenue

  27. Story Elements Model Estimation • Manually collected story surveys • 397 movies from estimation universe, 1995-2004 • All variables are de-meaned before regression • Story elements influence revenue independently of their correlation with genre

  28. Revenue Forecasting Factors Model Estimation • Sensible • Production cost • Talent (director, actor, writer, producer, …) • Studio • Rating • Season • Genre, story elements, demographics • Run time • Interactions (teams repeat, stars specialize by genre, …) • Excluded to prevent look-ahead • Advertising expense • Opening screens • Prediction markets (hsx.com)

  29. Comparison of Models Model Estimation 1. “No one knows anything” • Average revenue of past movies in estimation universe • EWMA with 5 year half life 2. Average over comparables in one of roughly 60 clusters • Production cost • Talent • Rating • Season • Genre 3. Procinea revenue model • All models updated monthly using realistic lags • Out of sample test in estimation universe

  30. Distribution of Dollar Forecast Error Model Estimation Movies released 1997 to 2004

  31. Valuation Ratio Strategy Results • Decision rule for each movie • Given attributes at greenlight, model predicts total revenue • Total revenue is divided among channels according to historical fractions • Channel revenue is scheduled according to historical time envelopes • Value is estimated as the present value of these cash flows at a fixed required rate • Project is accepted if value exceeds fully loaded production cost, including a cost-dependent estimate of P&A • 141 movies selected from 1997 to 2004 • 24% of target universe

  32. Movies Selected by Model Have Superior ROI Strategy Results

  33. Simulated Implementation, 2000-2004 Releases Strategy Results • Model can inform a strategy • Valuation and risk assessment are used to negotiate with studio • We simulate performance of hypothetical investor • Starting November 1998, each month at most one movie is chosen from those selected by the model, if any • Time from initial investment to release is stochastic 14-18 months • Continue until $1b total investment in $2b of production and P&A • Funded with $400m equity, remainder debt • Diversified, multi-studio portfolio of 15-20 films created in 2 years • Timing of cash flows assumed consistent with historical experience • Simulation pays studio distribution fee and participations • Repeat for subsequent inception dates • Fundamental independence of movie projects simplifies portfolio construction

  34. Distribution of Internal Rate of Return Strategy Results • 1% of simulations give negative IRR

  35. New Territory Extensions • Artistic and Intellectual Property • “The sheer volume of intellectual property worldwide is staggering,” B orod (2005), p.65 • Patents • Intellectual Ventures, Ocean Tomo, RIM vs. NTP, LabCorp vs. Metabolite • Other areas • Pharmaceuticals, video games • Non-linear models and interactions • Not a significant area of research in mainstream finance • Techniques suggest that useful progress could be made beyond standard linear factor technology

  36. Nobody Knows Anything? Ha! Summary • Forecasting revenue is not futile • Movie attributes can be used as a basis for an active investment strategy • Story elements do help • There is lots we don’t know • Models of compensation and advertising budget • Really understanding the role of stars • The impact of managing the contracting process • Sequels, animations, foreign movies, Bollywood,… • Thoughts for the future • Structure of the industry • organization, production process, who should benefit from movies, technology, etc • Art or science?