Blockbuster Software Titles and

Document Sample
Blockbuster Software Titles and Powered By Docstoc
					     Blockbuster Software Titles and Indirect Network
        Effects in the Home Video Game Industry

                                      Raymond Lee

                                    January 16, 2009


This paper investigates the indirect network effects in hardware/software system markets.
It is argued in the previous literature that the sales of the hardware is highly affected by the
availability of compatible software products. A majority of the previous literature has shown

that software quantity is a major factor influencing hardware demand. By doing so, the effect
on the hardware demand is assumed to be equal across software titles. (Nair, Chintagunta,
and Dub´ 2004) In this paper, I use a static model in which I let individual software titles

vary in how much they affect hardware demand. Different from the literature that also does

this(Lee 2008), I allow software titles to be substitutes(Derdenger 2008). I show that for the

6th generation video games and consoles that I use in my empirical analysis, one blockbuster
game can impact sales of other games on the same console. Then I test for my data that

high quality software titles have a bigger impact on hardware demand then do low quality
software titles. Further, I argue that a hardware platform needs a blockbuster title to attract


1    Introduction

Many high tech industries provide products that are usable in conjunction with a compatible
platform. These are referred to as hardware/software markets. Examples of such systems

are video console/game, mobile phone/service provider, and computer OS/software. The
consumer demand for platforms depend on the software compatible with the hardware. Thus

hardware suppliers need to form a good pool of compatible software in order to attract more

consumers. Having a good pool of software includes meeting various needs of consumers as
well as attracting mass demand with high quality software. In this paper, I examine the US

video game market to discuss the importance of high quality software.

The importance of software variety for hardware platform markets has been investigated in
the previous literature. Nair, Chintagunta, and Dub´ (2004) and Clements and Ohashi(2005)
show the importance of having a competitive number of compatible software titles. In

the fifth generation video game market(1995-2002), Sony, although a new entrant to the
market, penetrated the market by securing a wide variety of software titles so that they

could market PlayStation to a wider consumer base. Sony eventually became a big player
in the market. Sega, in contrast, failed to provide enough software variety and dropped out
of the market. While the number of software titles has an important impact on hardware

demand, the importance of having some specific high quality software titles is becoming

more and more important. The importance of a high quality blockbuster title appeared in
the sixth generation video game market(2000-2006).

Sony maintained its market power in this generation through its early entry of PlayStation 2.

PlayStation 2 provided backward compatibility of PlayStation games and also attracted game

developers to form a wide variety of software. Sony’s main competitors in the generation are
Microsoft’s XBOX, and Nintendo’s GameCube which were launched a year after PlayStation

2. XBOX had features superior to PlayStation 2, but struggled to gain market share because

it was not able to attract game developers. However, the launch of Microsoft’s own developed

and exclusive title Halo and its sequel Halo 2 was able to attract consumers to XBOX.
By October 2006, about 29.8% and 36.8% of XBOX owners had bought Halo and Halo 2
respectively. GameCube lacked such a blockbuster title, and as a result, gained a lower share

than XBOX. This is also partly because GameCube was targeted only to loyal and young
consumers. Lee(2008) and Derdenger(2008) examine this generation and provide frameworks

that allow software titles vary in how much they affect hardware demand.

In the following, I will describe some aspects of the data used, and predict demand systems

for hardware and software markets of the sixth generation video game market. This paper

differs from Lee(2008) in that I assume software titles to be substitutes within platform.
Unlike Derdenger(2008) where software utility is calculated directly from observed data
without estimation, I estimate software demand as well. This allows us to separate how

software titles affect hardware demand from other factors such as seasonal effect. I also
conduct counterfactual experiments to investigate the effect of blockbuster software title.

2    Data Description

The data consists of monthly sales and average prices of hardware and software that is

provided NPD group. While hardware sales and prices, and software sales data are available
from January 2001 to October 2006, software price data are only available from May 2004 to

October 2006. Thus I will use data from only this period. Software attributes data including

GameSpot review score, critic average review score, user average review score, title release
date, ESRB rating, and genre are gathered from GameSpot’s web site,

The collected review scores will be used as a proxy for consumers’ software quality perception.
I focus on the three main platforms, GameCube, PlayStation 2, and XBOX. From the data

described above we have three platforms and 1596 software titles with full information over

                                     Price($)            Unit Sales
                  Age of title    Mean        SD       Mean          SD
                             0   38.495 11.542     53599.504 159168.553
                             1   37.546 11.583     45722.028 99955.724
                             2   36.589 11.822     21708.663 57194.179
                             3   35.019 11.723     17255.550 35046.699
                             4   32.995 11.370     11612.810 41441.099
                             5   30.762 11.170      8508.927 17038.846
                             6   28.837 10.761      7131.245 12702.645
                             7   26.965 10.449      6293.121    9628.653
                             8   25.644 10.394      6088.471 13942.524
                             9   24.470 10.220      6157.226    9236.668
                           10    23.239 9.835       5534.699    9053.562
                           11    22.131 9.140       4766.507    7949.223
                           12    21.119 8.411       4993.864    9348.809
                           13    20.400 7.926       6405.855 14836.135
                           14    19.981 7.812       4845.288 10767.053

           Table 1: Mean prices and sales of software titles across age of game.

the period of May 2004 to October 2006.

Table 1 shows the mean of price and sales over all games by age. We can see there is a general
decreasing trend in both price and sales. This is due to market saturation, coming from the
fact that consumers purchase a software titles only once and tend to exit from the potential

market of that specific title.(Nair 2007) A similar pattern is observed in the hardware price

and sales except that hardware prices drop interruptedly at certain periods when hardware
providers cut store tagged price. Seasonality, like in many other markets is evident in the

data. Figure 1 shows that the total sales over the period of both hardware and software are

significantly high in December.

Sales of software titles are concentrated with a few successful blockbuster games. The top 14
games, which is only 0.58% of the total number of games, accounts for 10% of the total sales

of software titles. This phenomenon is most apparent for XBOX. The biggest blockbuster

of XBOX, Halo and its sequel Halo 2 together account for 8% of the game sales compatible

                                                                                       6 x106 (hw)
         6 x107 (sw)





                       Feb      Apr         Jun         Aug         Oct         Dec


                              Figure 1: Total sales per month.

with XBOX. This motivates the existence of blockbuster effect in software/hardware system
markets. Such high quality software titles negatively affect sales of competitors and pos-
itively affect sales of the corresponding platform. A reduced form analysis examining the
competitive aspect between software titles is provided in section 3.1.1.

The May 2004 to October 2006 period includes introduction of two major blockbuster titles.
Grand Theft Auto: San Andreas for PlayStation 2 in October 2004 and Halo 2 for XBOX
in November 2004. These two software titles record unit sales over two times the next
best exclusive title, and over five times Paper Mario: Thousand, the biggest introduction

for GameCube in the period. The period of the data having dominant and compatible

blockbusters for only two of the platforms sheds the potential of identifying the blockbuster

3       Demand Models

3.1     Video Game Demand

For the video game demand, the consumer’s utility for game k in month t given that having
purchased console j is

                         uSW = βP Pkt + Xkt β SW + ξkt + εikt
                                SW SW    SW

Pkt is the price of game k, Xkt is a vector of game characteristics, ξkt is the mean unobserved

characteristics of software k common over individuals at time t, and εikt is the individual
deviation from the mean utility. Note that the outside good’s utility is set to have mean

Popular video games often become published in alternative editions. Also, games requiring

additional accessories for game play are available as an alternative bundle with the accessory.
The share of these software titles are likely to be under-counted because the sales get split
across these alternative editions. For this reason, in estimation, I add the sales of all available
packages for a software title, and use the average of the package prices weighted by sales.

Game characteristics include game specific fixed effects for all software titles which enables

me to capture all time invariant software characteristics, and the CPI adjusted average sales

price. Platform fixed effects are included to let identical games have different sales levels on

each platform, because the games are facing different consumers. I also include a nonlinear
function of game age in order to capture market saturation and other temporal phenomena

such as nonlinear patterns associated with a new release. There is a possibility of game age

reflecting a partially positive effect on demand such as awareness of word-of-mouth, but this
is likely dominated by the saturation effect given the general trend of decreasing price and


I assume that εikt , the individual taste follows an independent type-I extreme-value distri-
bution. Then the market share of software k is given by

                                   SW SW      SW
                              exp(βP Pkt + Xkt β SW + ξkt )
                    skt =               SW SW     SW
                          1 + l∈Kj exp(βP Plt + Xlt β SW + ξlt )

Consequently, following Berry(1994), we get a linear regression model as

                        log            SW SW    SW
                                    = βP Pkt + Xkt β SW + ξkt                              (3)

I am assuming here that consumers make discrete choice so that they purchase only one

software title in a period. This simplifies the definition of s0t and allows us to write equation

In defining the market size, mt , I use the installed base of platform j at time t. Thus I am
assuming that a consumer can repurchase the software title with some probability. This is
a strong assumption and requires relaxation due to the fact that video games are durable
goods. However, inclusion of the nonlinear function of video game age will partly capture
the relative change in market size across the various software titles. The demand for software

product k is

                                         qkt = mt skt

where mt is the market size at time t.

3.1.1    Competitive Aspect of the video game market

To assert the competitive nature of video games, I briefly examine the effect of a blockbuster

game introduction. In particular, I consider the effect of the introduction of the best-selling

Playstation 2 action game Grand Theft Auto: San Andreas. I examine how the sales and

prices of software titles competing to the blockbuster are affected. As a reduced form ap-
proach, I do OLS regressions of sales and price of software titles sharing genre with Grand
Theft Auto: San Andreas on a dummy variable indicating that a title is introduced at the

same period October 2004, as the blockbuster title. This dummy variable will capture the
effect of the introduction of Grand Theft Auto: San Andreas on competing software titles. I

also control for age and seasonal effects. I repeat same for the top selling sports game Mad-

den NFL 2005 that was introduced at August 2004. The results are given in table 2. We

can see that both the sales and price of competitors of Grand Theft Auto: San Andreas are

lower than the overall average. For Madden NFL 2005, competitors have sales significantly
lower with no significant difference in price. Thus we can say that competitors of a high

quality software title are negatively affected if introduced in the same period as the high
quality title. The control variables all give coefficients in the expected direction.

                                      Unit Sales                               Price
                           Estimate   Std. Error   t-statistic   Estimate   Std. Error   t-statistic
           Intercept      15867.965     544.2460      29.1559     28.0120       0.1768    158.4106
                Age        -432.985      19.5244     -22.1766     -0.3272       0.0063     -51.5772
            Holiday       18010.980    1175.7239      15.3191     -0.0743       0.3820      -0.1944
 Entry on Oct. 2004       -7371.364    1690.3899      -4.3607     -3.6553       0.5492      -6.6554

                                   Unit Sales                                  Price
                          Estimate Std. Error      t-statistic   Estimate   Std. Error   t-statistic
           Intercept    15734.8482   363.7391         43.2586     28.5455       0.1066    267.7595
                Age      -401.0932    11.4584        -35.0043     -0.3193       0.0034     -95.0739
            Holiday     14405.2604   757.3136         19.0215      0.2281       0.2220       1.0275
 Entry on Aug. 2004     -6286.6267 2709.3624          -2.3203      0.7427       0.7941       0.9352

Table 2: OLS regressions to test changes in sales and price of games introduced at the same
period as a blockbuster game. Top is for action games against Grand Theft Auto: San
Andreas, bottom is for sports games against Madden NFL 2005.

3.2    Game Console Demand

For the game console demand, the consumer’s utility for console j in month t is

                          HW HW    HW
                   uHW = βP Pjt + Xjt β HW + αSW Ujt + ζjt +
                    ijt                                                    ijt                      (4)

                     Ujt = log(              u
                                         exp(˜kt ))                                                 (5)

Pjt is the price of console j, Xjt is a vector of console characteristics, ζjt is the mean unob-

served characteristics of console j at time t common over individuals, and       ijt   is the individual

deviation from the mean utility. Ujt is a measure of utility from software titles available at
time t. Unlike Derdenger(2008), where software demand estimation is omitted and the ob-
served shares are used to calculate software utility, I use the fitted utility of individual games

from the previous software demand system. By doing so, I am able to separate the effect of
software availability from the holiday effect. I use ukt which is the expectation of ukt with
the seasonal effect eliminated. Without the separation, Ujt becomes highly affected by the
seasonal effect, making it hard to distinguish the software effect from seasonal effect. While

any additional software will give a positive effect on console demand, software utility are

negative for all games because it is normalized so that the outside good gives zero utility. I
take the exponential of software utility and sum up the exponential of utility available for

each console each month to get the software availability measure, and again take the loga-

rithm to get Ujt . Since I use the software utility through the demand system in the previous
section, software titles have heterogeneous and time variant effect on the hardware demand.

The console characteristics such as processor speed, CPU bits, or graphics quality do not

change over time, so I use console specific dummy variables to capture the time invariant
characteristic effects of each hardware system. I also include console age and seasonal effect

variables to capture time dependent effects.

As in the video game demand, I assume       ijt   to follow an independent type-I extreme-value

distribution. This leads to the following model similar to the video game demand model.

                  log            HW HW    HW
                              = βP Pjt + Xjt β HW + αSW Ujt + ζjt                            (6)

Skt is the share of hardware console j in time t. The demand for hardware console j in time
t is

                                       Qjt = Mt Skt                                          (7)

where Mt is the market size at time t. I follow Derdenger(2008) to determine the market size,

where the predicted initial market size is M = 78, 354, 700 and the market size at time t,
Mt , is Mt = M − (cumulative console sales). By this I implicitly assume that any consumer

exits the market along with purchase.

3.3    Estimation

I estimated the game demand equation using a conditional moment with ξkt , the unobserved
characteristics. The condition is E(ξ|Z) = 0, where Z is a set of instruments uncorrelated to
ξ. For software price, the use of commonly used instruments is likely inappropriate because

marginal cost is constant, regional data is not available, and software attributes explain little

of the pricing process.(Nair 2007) In belief that games at similar quality levels and being

in the same genre will share similar pricing patterns, I use game software development PPI

interacted with GameSpot review score and genre fixed effects.

In estimation for the console market, I again use instrumental variables to handle the corre-
lation of variables with the demand shock. Instruments used for price, following Liu(2007),

are PPI for computers and computer storage devices. For the software availability, I use the

logarithm of the number of compatible software available in the market for each platform

each period.

4     Results

4.1    Software demand

I first examine the game specific fixed effects. Table 3 shows an OLS regression of the game

specific fixed effects on the game characteristics, genre and ESRB rating. This shows how
much of the software demand is explained by the time invariant software attributes. None
of the genres are significantly higher than the base genre action, while Adventure, Flight,

Racing, Sports, and Strategy are lower. For the ESRB rating, we can see that Everyone 10+
and Mature have higher effects on the game specific fixed effects. GameSpot review score has
a significantly positive effect providing evidence that higher quality leads to higher demand.
These fixed effects only account for an R2 of 0.1355 on the game specific fixed effects.

Table 4 shows the software demand coefficient estimates from OLS and GMM. All estimates

are significant in the intended direction. Platform fixed effects are included to control for
each market. Since Playstation 2 has more games, relative demand of software titles is mea-
sured lower than the two other platform markets. We can see that the price coefficient is

significantly negative as expected. The nonlinear function of game age significantly cap-

tures the decreasing trend of a software title over time agreeing with the market saturation
argument.(Nair 2007) Seasonal effect is shown to be evidently strong.

                                    Estimate Std. Error   t-statistic
                       Intercept    -12.1929     0.2713     -44.9347
                      Adventure      -0.7288     0.2321      -3.1403
                         Arcade       0.2619     0.4845       0.5406
                      Children’s      0.1855     0.8291       0.2237
                          Family      0.2139     0.2462       0.8686
                        Fighting      0.1343     0.2218       0.6053
                           Flight    -0.7802     0.3710      -2.1032
                           Other      0.5422     0.8238       0.6582
                          Racing     -0.6928     0.1840      -3.7648
                    Role-Playing     -0.2217     0.2000      -1.1087
                         Shooter     -0.1446     0.1876      -0.7709
                          Sports     -0.8006     0.1717      -4.6619
                        Strategy     -0.7824     0.3064      -2.5538
                   Everyone 10+       2.2951     0.2704       8.4865
                         Mature       0.8121     0.1737       4.6763
                            Teen      0.1675     0.1358       1.2329
                      GameSpot        0.3080     0.0345       8.9322

Table 3: OLS of game fixed effects on Genre and ESRB. Base is Action and Everyone.
GameSpot is for GameSpot review score.

                             OLS                               2SLS
                 Estimate Std. Error t-statistic   Estimate Std. Error    t-statistic
    GameCube       2.3304     0.0406    57.4490      3.1124     0.0892       34.8740
          PS2      2.1839     0.0396    55.1816      2.9655     0.0888       33.4084
        XBOX       2.2896     0.0391    58.5038      3.0431     0.0860       35.3769
         Price    -0.0186     0.0006   -29.6849     -0.0367     0.0019      -18.8838
           Age    -0.0112     0.0028    -4.0119     -0.0140     0.0028       -4.9680
         Age2      0.0000     0.0000     0.0620      0.0001     0.0000        1.9287
   log(Age+1)     -0.5846     0.0227   -25.7906     -0.7179     0.0265      -27.1001
       Holiday     2.1354     0.0200 106.6941        2.1381     0.0201     106.3268

Table 4: Software demand coefficient estimates. Holiday is a dummy variable for December.
Game specific fixed effects not reported.

4.2    Hardware demand

Table 5 reports the coefficient estimates for the console demand system. All estimates
are significant in the intended direction. PlayStation 2 has the highest demand, followed
by XBOX. Price has a significantly negative estimate. The software utility coefficient is

significantly positive showing that the indirect utility coming from software availability is
a big factor in hardware demand. The seasonal effect is also significant. The combination
of age effect is negative since the ages of consoles are over 30 months in the data set. This

again can be explained by market saturation.

                                OLS                               2SLS
                  Estimate   Std. Error t-statistic   Estimate Std. Error t-statistic
   GameCube        -1.2427       1.1213    -1.1082     -0.6590     1.0102    -0.6523
         PS2        3.5488       1.3500     2.6288      6.7950     1.3882     4.8947
       XBOX         2.3701       1.3609     1.7416      5.6995     1.4053     4.0556
        Price      -0.0548       0.0093    -5.9073     -0.0921     0.0115    -8.0375
     Software       0.1400       0.1857     0.7542      0.4823     0.1860     2.5934
      Holiday       1.9715       0.2316     8.5116      2.0262     0.2072     9.7788
          Age       0.0694       0.0584     1.1882      0.2166     0.0608     3.5611
         Age2      -0.0012       0.0006    -1.8565     -0.0029     0.0007    -4.2880

          Table 5: Hardware demand coefficient estimates from OLS and 2SLS.

5     Counterfactual Experiment

In this section, I use the demand estimates to conduct counterfactual experiments that
capture the heterogeneous effect of software titles, especially, blockbuster games. I do this by

switching software attributes, calculating software utility using demand parameter estimates,
and forming the new software utility index that affects hardware demand. After retrieving

new hardware share, I multiply the market size to get new hardware sales. This in turn

updates the installed base for the next period. Thus I observe the effect of software attribute
change on installed base. Note that this is only partially counterfactual because I take

software title prices unchanged from the original data, while it is expected to be correlated

with other newly adjusted software attributes.

5.1    Equal fixed effects

The software attribute that I experiment with is the game specific fixed effects obtained from

the software demand system. Table 6 gives descriptive statistics of the fixed effects. First
I fix all of the game fixed effects for each platform at an equal level. This will upgrade the
effect of games at the low end, and downgrade the effect of those at the high end. Figure

2 shows how installed base changes over time when all game fixed effects are fixed at mean

level. We can see that the sales decreases dramatically. The decreased sales are 3,351,847,
9,272,720, and 3,853,845 units respectively for GameCube, PlayStation 2, and XBOX. This

implies that the utility loss from high end titles is higher than the gain from low end titles.

                           Min. 1st Quar.     Median    Mean      3rd Quar. Max.
          GameCube       -15.47     -11.26     -9.578 -9.584          -7.954 -4.762
               PS2       -15.85     -11.85    -10.070 -10.220         -8.534 -4.607
             XBOX        -14.79     -10.88     -9.493 -9.555          -8.212 -4.607

               Table 6: Descriptive statistics of the game specific fixed effects

I manually searched for fixed values, -8.28, -8.91, -8.66, that give hardware installed base

close to reality. These are about the 31st, 31st, and 33th percentiles of the game specific
fixed effects within each platforms. This supports the argument that high quality software
have higher effects on hardware demand. The results are shown in Figure 3.

5.2    Blockbuster effect

The previous section shows some evidence that high quality titles have larger effects on the

hardware sales. In this section, I investigate the effect of specific blockbuster titles. I examine

the two major blockbuster introductions in the period, Grand Theft Auto: San Andreas for
PS2, and Halo 2 for XBOX. Both of these titles are sequels of blockbuster games. Grand

Theft Auto: San Andreas and Halo 2 sold 6,510,244 units and 5,323,384 units respectively
during the time period. These correspond to 18.7% and 36.8% of the installed base of PS2
and XBOX.

First I examine the result of downgrading the quality of Grand Theft Auto: San Andreas of

PS2. I do this by setting the title’s game specific fixed effect at the mean level. Figure 4 shows
the result. We can see a decrease in the PS2 installed base that XBOX and GameCube takes
away. PS2 loses 377,104 unit sales with Grand Theft Auto: San Andreas absent. XBOX

gains 26,034 unit sales and GameCube attracts 41,111 consumers.

Now I examine the opposite case, Halo 2 being downgraded. Figure 5 shows the result.

XBOX loses 472,886 unit sales. PS2 gains 132,507 unit sales and GameCube attracts 42191

The loss of unit sales for PlayStation 2 and XBOX in the two experiments account for 3.31%
and 8.17% of the total reality sales. Considering the loss is from losing a single software

title, this evidently shows the strong impact of blockbuster titles.

Now I try adding median quality software titles to XBOX after degrading the quality of

Halo 2, to see how many titles are needed to overcome the loss of a blockbuster. Manual
search using the model suggests that hardware sales of an equal level can be reached with
22-23 more median level software titles. The total number of software titles for XBOX in

the whole data set is 800, and the number available at the introduction of Halo 2 is 496.
This means that XBOX needs to introduce 7.57% more games than reality in this period, to

reach the level of real hardware sales according to the model.

6    Conclusion

In this paper we examined indirect networks effect in software/hardware system markets.
We find by using GameSpot review scores as a proxy, that software quality leads to higher

demand of a software title. It is also discussed that the software titles market has a com-
petitive nature in that an introduction of a blockbuster title negatively affects sales and

prices of competitors. High software quality leads to higher hardware demand by raising
software availability related utility. Counterfactual experiments show that high quality, or
blockbuster, software titles have stronger impacts on the hardware demand than average
titles. The implication of this is that it is more important having software with dominant

impact than having many mediocre software titles, for a hardware platform to attract more

The used model has limitations in sense that it does not account for consumer heterogeneity.

An extended model with random coefficients or latent-class will be needed to overcome this
limitation. Also, the software market size definition does not suitably account for market

saturation. Improvement on these concerns are left for future research.

Investigating the effect of having special editions is another issue to be studied in the future.

Special, limited, or collector’s editions normally have about the same quality for the software

itself, but comes with good packages or some small exclusive additions to the software title.
These types of editions are usually available for blockbuster titles, and the proportion of
sales of the edition likely depends more on the quality than the additional price. Future

research on the special edition effect will likely find evidence for additional profits in the
strategy. Setting up a detailed framework on the competitive software market incorporating

effects of software ownership will be a direction for this.


 • Berry, Steven T. (1994). ”Estimating Discrete-Choice Models of Product Differen-

   tiation,” RAND Journal of Economics, The RAND Corporation, vol. 25(2), pages

   242-262, Summer

 • Clements, Matthew T. and Ohashi, Hiroshi,(2005) ”Indirect Network Effects and the

   Product Cycle: Video Games in the U.S.”, 1994-2002. Journal of Industrial Economics,

   Vol. 53, No. 4, pp. 515-542,

 • Derdenger, T. (2008), ”Quantifying the Effects of Vertical Integration on Platform

   Pricing: Application to the Video Game Industry.” mimeo

 • Lee, R.,(2008) ”Vertical Integration and Exclusivity in Platform and Two-Sided Mar-
   kets” mimeo

 • Liu, H.,(2007) ”Dynamics of Pricing in the Video Game Console Market: Skimming
   or Penetration?” mimeo

 • Nair, H., Chintagunta, P. and Dub´, J-P. (2004). ”Empirical Analysis of Indirect
   Network Effects in the Market for Personal Digital Assistants,” Quantitative Marketing

   and Economics 2(1), 23-58

 • Nair, H., (2006) ”Intertemporal Price Discrimination with Forward-Looking Consumers:
   Application to the US Market for Console Video-Games.” Stanford University Gradu-

   ate School of Business Research Paper No. 1947

                        GameCube                                           PS2                                            XBOX












        Jun04   Dec04    Jun05   Dec05   Jun06           Jun04   Dec04   Jun05   Dec05   Jun06           Jun04   Dec04   Jun05   Dec05   Jun06

Figure 2: Simulated installed base scenario 1: All games fixed at mean level. Solid and
dotted lines are actual and simulated installed bases respectively.

                        GameCube                                           PS2                                           XBOX













        Jun04   Dec04    Jun05   Dec05   Jun06           Jun04   Dec04   Jun05   Dec05   Jun06           Jun04   Dec04   Jun05   Dec05   Jun06

Figure 3: Simulated installed base scenario 2: Hardware sales reaches real sales level when
games fixed at 30 33th percentile quality. Solid and dotted lines are actual and simulated
installed bases respectively.

                        GameCube                                           PS2                                            XBOX












        Jun04   Dec04    Jun05   Dec05   Jun06           Jun04   Dec04   Jun05   Dec05   Jun06           Jun04   Dec04   Jun05   Dec05   Jun06

Figure 4: Simulated installed base scenario 3: PlayStation 2’s hit Grand Theft Auto: San
Andreas degraded. Solid and dotted lines are actual and simulated installed bases respec-

                    GameCube                                               PS2                                           XBOX













        Jun04   Dec04    Jun05   Dec05   Jun06           Jun04   Dec04   Jun05   Dec05   Jun06           Jun04   Dec04   Jun05   Dec05   Jun06

Figure 5: Simulated installed base scenario 4: XBOX’s hit Halo 2 degraded. Solid and
dotted lines are actual and simulated installed bases respectively.


Shared By: