Slide1: An Empirical Analysis of Indirect Network Effects in the Home Video Game Market Wei-Min Hu Shenzhen Graduate School of Business of Peking University James E. Prieger Pepperdine University School of Public Policy April 2007


Indirect Network Effects: Indirect Network Effects Hardware Sales Software Variety Software Sales and Profit DVD(CD), MP3, ATM, PDA, Video Games


Why the Video Game Market?: Why the Video Game Market? Data are available, have been used in other studies. Yearly sales in the video game industry are nearly $10 billion ( > Hollywood’s box office) The average gamer spends more than triple the amount of time playing video games each week than time spent in exercise, reading, community, religious, and creative activities… combined!


Our Study Examines: Our Study Examines Relative impact of console price and game variety on market share Is there an applications barrier to entry? How well our model predicts out-of-sample. Are these structural models generalizable? Both are important, but the variety effect has become less important relative to the price effect There is little evidence for an applications barrier to entry Predicts Japanese video market well, does not match other predictions for U.S. market


Literature Review: Literature Review Theoretical Chou and Shy (1990) Church and Gandal (1992; 1994) Nair et. al. (2004) Empirical Gandal, Kende, and Rob (2000) –Compact Disc Player Dranove and Gandal (2003)—DVD and Divx Nair, et. al. (2004)—Personal Digital Assistants Clements and Ohashi (2005)—Video Games


Evolution of Video Game Consoles: Evolution of Video Game Consoles 1st Generation (late 1970s) ATARI PONG ATARI 2600


Home Video Games: Home Video Games MATTEL INTELLIVISION COLECO ATARI 5200 2nd Generation (early 1980s)


Home Video Games: Home Video Games 3rd Generation (mid-late 1980s) ATARI 7080 SEGA MASTER NES


Home Video Games: 4th Generation (early 1990s) Home Video Games SEGA GENESIS PHILIPS CDI SNES


Home Video Games: 5th Generation (1995-2000) Home Video Games PlayStation I SEGA SATURN ATARI JAGUAR N64


Home Video Games: Home Video Games 6th Generation (2000-2005) PlayStation2 XBOX GameCube Dreamcast


The Market for 6th Generation Home Video Games : The Market for 6th Generation Home Video Games Sony’s PS2 is the market leader, both in console sales and in game titles Was introduced a year earlier than XBOX and GameCube Has slower processor, less RAM PS2 and XBOX sell for about same price GameCube is cheaper, has smallest mkt share PS2 has largest proportion of exclusive game titles


Slide13: Market Share of Video Game Consoles


Empirical Model: Empirical Model Based on the static structural model of Nair et al. (2004) Hardware adoption (demand) Software supply


Empirical Model: Empirical Model Hardware demand: Logit discrete choice cj : console dummies and year dummies dt : holiday dummies Why not use nested logit? Do not have good instruments for within group share


Empirical Model: Empirical Model Software Supply


Endogeneity and Instruments: Endogeneity and Instruments Endogeneity in hardware adoption: Pjt : a shock to brand image (in xjt ) will also affect prices Njt : a shock to xjt-1 + autocorrelation in the HW supply equation + indirect network effects in SW supply equation => Njt correlated with xjt Instruments: Pjt: Japanese retail console price (Nikkei News) and current Japanese–US exchange rates Njt: Japanese game variety (Famitsu) Console age, age2 Will carefully check strength and validity of instruments


Data: Data Potential market size: households with at least one television Statistical Abstract of the United States (Census Bureau, 2004-2005) Console sales (monthly, Mar 2002-Dec 2004): NPD fun group Price (adjusted by CPI): 7 major retail chains Game titles: Gamespot.com


Results: Console Demand: Results: Console Demand Hausman test rejects the OLS in favor of the IV


Slide20: Estimation Results: Software Supply Model


Strategic Tradeoffs Between Lowering Price and Enhancing Game Supply: Strategic Tradeoffs Between Lowering Price and Enhancing Game Supply -es/ep: %DP that increases demand as much as a 1% increase in game variety Average = 0.39, but decreasing over time


Is There An Applications Barrier to Entry?: Is There An Applications Barrier to Entry? Barriers to entry based on software applications for a system received much discussion in Microsoft antitrust case Antitrust Concern: Can a console maker hinder entry by competing systems through exclusive games.


Do the Network Effects Differ for Exclusive Games?: Do the Network Effects Differ for Exclusive Games? Predatory conduct of the console maker would be more effective if unique games have strong indirect network Re-estimate the console demand model by splitting unique and non-unique games Results: Little network effect for unique games… …it all comes from the nonunique games.


Counterfactuals to Assess Applications Barriers to Entry: Counterfactuals to Assess Applications Barriers to Entry Another way to assess the feasibility of using unique games as a strategic weapon: counterfactual scenarios no firm has the ability to create unique games only the non-dominant firms have the ability to create unique games Outlawing exclusively provided games does not change PS2’s dominant status in console market share


Is There An Applications Barrier to Entry? Summary: Is There An Applications Barrier to Entry? Summary Unique games appear to be less effective than non-unique games in changing console demand Even outlawing exclusively provided games does not change PS2’s dominant status in console market share There may not be much reason for concern regarding applications barriers to entry in video game market


Out-of-Sample Prediction: Out-of-Sample Prediction Compare our estimation result to that of Clements and Ohashi (2005) 2002 is at the end of their sample, and the beginning of ours:


Out-of Sample Prediction: Compare US Data to Japan Data: Out-of Sample Prediction: Compare US Data to Japan Data Turn our data around and estimate a Japanese console demand equation Price and software elasticities are about the same So: model generalizes horizontally but not over time?


Conclusion: Conclusion Indirect network effects are present on both sides of the market We find little evidence of applications barriers to entry through creating exclusively provided games We find little generalizability of our results to previous markets Flip side: previous studies do not extrapolate well to 6th generation Policy makers cannot escape doing the case study anew?