Impact of Competition on Technology Adoption: An Apples-to-PCs Analysis NY Federal Reserve and BEA Nov 2010
Disclaimer The views expressed herein are our own and not necessarily those of the Bureau of Economic Analysis, the US Department of Commerce, the Federal Reserve Bank of New York, or the Federal Reserve System
Motivation What is the role of competition on technology adoption? We focus on personal computer industry Computer manufacturers adopt innovative intermediate inputs how does that adoption decision depend upon competition? We build on a literature examining the relationship between competition and innovation Schumpeter: competition impedes innovation because it dampens rewards from innovation. Arrow: competition drives innovation because firms need to distinguish their products from competing products. Important implications across a variety of fields, including economic growth and anti-trust.
Overview of the Paper Why analyze the personal computer industry: Two retail computer markets: 1 IBM-platform ( PC ) (many firms) 2 Apple-platform (one firm)
Overview of the Paper Document a number of stylized facts describing differences across 2 markets Use a model to examine the importance of market structure in explaining these differences Main result: Model predicts that competition drives faster rate of tech adoption, in support of Arrow.
NPD Techworld scanner data Monthly unit sales and revenue by computer-model Impute price from sales and revenue data Specifications of each computer-model (e.g. laptop, screen-size, and RAM). November 2001 to April 2009 Collected from retail stores misses direct manufacturer-consumer sales (e.g. Dell).
Metafact s TUP survey data Four annual surveys, 2001 and 2004 Collects info. on use of IT products at home We use: household income, primary computer, and price paid.
Summary of Stylized Facts PCs Apple 1. Competitive market Monopolized Market 2. Frequent product entry Infrequent product entry 3. 4 month product cycle 8 month product cycle 4. Sales bunched at introduction Sales spread over product cycle 5. Prices decline over product cycle Prices are flat over product cycle 6. Wide Income Distribution Narrow and High Income Distribution
%"!!!""#$ %!!! $"!! $!!! %&'($ #"!! #!!! "!! $!!*&#!! $!!#&#! $!!$&#! $!!%&#! $!!'&#! $!!"&#! $!!(&#! $!!)&#! $!!*&#! %"!! )*+,'-. %!!! $"!! $!!! %&'($ #"!! #!!! "!! $!!*&#!! $!!#&#! $!!$&#! $!!%&#! $!!'&#! $!!"&#! $!!(&#! $!!)&#! $!!*&#!
%"!! 2$3#$445%.(6.&7 %"!!!""#$ %!!! %!!! $"!! $"!! $!!! $!!! %&'($ #"!! %&'($ #"!! #!!! #!!! "!! "!!! $!!#&#! $!!$&#! $!!%&#! $!!'&#! $!!"&#! $!!(&#! $!!)&#! $!!*&#! %"!! /*01 %!!!! $!!#&#! $!!$&#! $!!%&#! $!!'&#! $!!"&#! $!!(&#! $!!)&#! $!!*&#! %"!! )*+,'-. %!!! $"!! $"!! $!!! %&'($ #"!! $!!! %&'($ #"!! #!!! #!!! "!! "!!! $!!#&#! $!!$&#! $!!%&#! $!!'&#! $!!"&#! $!!(&#! $!!)&#! $!!*&#!! $!!#&#! $!!$&#! $!!%&#! $!!'&#! $!!"&#! $!!(&#! $!!)&#! $!!*&#!
Adoption of Intel CPUs 9 8 7 6 Age (Months) 5 4 3 2 1 0-1 2006m6 2006m10 2007m2 2007m6 2007m10 2008m2 2008m6 2008m10 2009m2 Hewlett Packard Toshiba Apple
Age of Newest CPUs (Notebook Computers) All CPUs Intel CPUs Avg. Max Avg. Max Hewlett Packard 0.8 3 1.1 3 Compaq 1.1 4 1.4 5 Toshiba 0.9 4 1.0 4 Apple 2.9 8 2.9 8 Gateway 1.5 5 1.5 5 Sony 1.5 7 1.5 7
Sales cdf over the Product Cycle by Brand 100% 90% 80% 70% 60% CDF 50% 40% 30% 20% 10% 0% 1 2 3 4 5 6 7 8 9 10 11 Months on Market PC 31 Days PC 1 Day Apple 31 Days Apple 1 Day
Price Declines over Product Cycle 102 100 98 96 94 Price Level 92 90 88 86 84 82 10% 20% 30% 40% 50% 60% 70% 80% 90% CDF PC Apple
Updating Dynamics: Frontier 15 inch 512MB Laptops 2400 2300 2200 2100 Chip: 3.0 GHz Pentium IV Display: 1280 x 800 Pixels HD: 60 GB Weight: 8.0 lb Chip: 1.5 GHz Pentium M Display: 1920 x 1200 Pixels HD: 80 GB Weight: 7.3 lb 2000 Price 1900 1800 1700 Chip: 1.5 GHz Centrino Display: 1280 x 800 Pixels HD: 60 GB Weight: 6.5 lb 1600 1500 Chip: 1.5 GHz Centrino Display: 1280 x 800 Pixels HD: 80 GB Weight: 6.5 lb Chip: 3.0 GHz Pentium IV Display: 1280 x 800 Pixels HD: 80 GB Weight: 7.8 lb 1400 2003m9 2003m11 2004m1 2004m3 2004m5 2004m7 2004m9 Compaq Hewlett Packard Sony Toshiba
Empirical exercise Big Question: What is driving the differences between two retail markets? Claim: Market Structure Empirical exercise: 1 Develop a vintage-capital model 2 Choose parameters to match model to PC prices and sales 3 Ceteris paribus, switch from competition to monopoly 4 Compare model s predictions on (price, sales) to data on Apple.
Model Environment Infinite horizon model N single product firms (no potential entrants) Computers are differentiated by vintage, ν (one dimension of differentiation) Each period, mass M of consumers arrive. They buy/not buy and leave. Consumers are differentiated by budget for computers, y B(κ 1, κ 2, b) over [a, b]. Demand system (Shaked and Sutton 1983,1984) Static problem buy a computer or outside option. Utility u ν (y p νt ) if purchase vintage ν (1) ǔ t y if purchase outside good (2)
Model Firm Single-product firms, i = 1, 2,..., N Beginning of period, decide on price, p νi t End of period, decide if update computer, d it = {0, 1} if update, then adopt latest technology and pay fixed cost φ > 0 Constant marginal cost c, same across all firms State variables s t = (ν 1, ν 2,..., ν N, ǔ t ) Given competitors strategies, firm i = 1, 2,..., N solves { V i (s t ) = max (p νi t c)q νi (p νi t, p ν i t; s t )+ (3) p νi t,d it [ ( β d it E[Vi (s t)] φ ) ]} + (1 d it )E[V i (s t )] (4)
Equilibrium Equilibrium concept is Markov perfect equilibrium Focus on stationary equilibrium sales and prices of a computer over product cycle are independent of time Assume that frontier computer quality and outside good grow at same rate Firm strategy in stationary equilibrium where the lowest-quality firm updates all other firms do not update in initial period N firms each produce a different computer vintage
Parameterization Computer quality frontier value is normalized at 10 Quality growth rate is fixed at 1.029 based on Moore s law ǔ t u t is fixed at 0.01 Income Cost Lower and upper income bounds: [a = 1, b] density of consumers, B(κ 1, κ 2 ). Market size, M > 0, marginal cost: mc 0
Parameterization algorithm Pick 5 parameters to match model s prediction of sales and prices over the product cycle to PC data: 1 price decline over product cycle 2 sales cdf over product cycle 3 product cycle length
Visual Display of Equilibrium Density of consumers 1200 1000 800 600 400 200 Purchasers of vintage t-3 Purchasers of vintage t-2 Purchasers of vintage t-1 Vintage Marketshare Markup t 47.8 18.4 t-1 33.3 4.8 t-2 15.2 2.2 t-3 3.7 0.5 Purchasers of vintage t 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Consumer budget
Model versus Data: Prices over the Product Cycle -0.02 PC data Competition model -0.04-0.06 Price decline (percent) -0.08-0.1-0.12 Month 1-2 1-3 1-4 Avg Data -0.9-14.7-16.8-13.5 Model -11.5-14.5-15.1-13.7-0.14-0.16-0.18 20 30 40 50 60 70 80 90 100 CDF
Model versus Data: Sales cdf over the Product Cycle )""# ("# '"# &"# %"#!"# $"#!"# +"# *"# Months 1 2 3 4 Data-lower bound 0.188 0.500 0.804 1 Model 0.478 0.811 0.963 1 Data-upper bound 0.500 0.804 1 1 )"# "# ) * +! $%&'()*%&*$+,-.',-./012-345 677/012-345 8-5/9
Model versus Data: Sales pdf over the Product Cycle 0.6 0.5 0.4 Percent of total sales 0.3 0.2 0.1 0 1 2 3 4 Months on market Lower bound Model Upper bound
Time-series of price and sales, data and model Price decline, relative to first month Month 1-2 1-3 1-4 average Data -9.0-14.7-16.8-13.5 Model -11.5-14.5-15.1-13.7 Sales cdf Month 1 2 3 4 Data-lower bound 0.188 0.500 0.804 1 Model 0.478 0.811 0.963 1 Data-upper bound 0.500 0.804 1 1
Single product monopoly case Faces same environment and same choices of price/updating same set of consumers same growth in product quality same marginal cost Big difference from competitive case no market-specific obsolescence Solve for optimal prices and replacement cycle Compare profitability of different replacement cycles over 500 periods. We select fixed cost to match length of product cycle with upper bound constraint, φ < Π C i.e. ensures monopoly and competitive cases are consistent
Model versus Data: Prices over the Product Cycle 0.02 0 0.02 0.04 Price decline (percent) 0.06 0.08 0.1 0.12 0.14 0.16 0.18 20 30 40 50 60 70 80 90 100 CDF Apple data PC data Monopolist model Competition model
Model versus Data: Sales cdf over the Product Cycle 100% 90% 80% 70% 60% CDF 50% 40% 30% 20% 10% 0% 1 2 3 4 5 6 7 8 Months on Market Competition lower bond Competition upper bound Competition model Monopolist lower bound Monopolist upper bound Monopolist model
Model versus Data: Sales pdf over the Product Cycle 0.17 0.15 Percent of total sales 0.13 0.11 0.09 0.07 1 2 3 4 5 6 7 8 Months on market Data Model
Monopolist: Data versus Model Price decline, relative to first month 1-2 1-3 1-4 1-5 1-6 1-7 1-8 average D -0.06-0.07-0.26-0.51-0.83-1.21-1.65-0.64 M -0.03-0.05-0.08-0.11-0.14-0.17-0.20-0.11 Sales cdf 1 2 3 4 5 6 7 8 D-low 0.09 0.25 0.42 0.57 0.71 0.83 0.92 1 M 0.13 0.25 0.38 0.50 0.63 0.75 0.88 1 D-up 0.25 0.42 0.57 0.71 0.83 0.92 1 1
What do we learn? A simple vintage capital model can capture main differences between PC and Apple by only changing market structure assumption. Calibrated model matches well PC facts Model s out-of-sample prediction for Apple also closely matches Apple data. The good fit of the model to Apple s stylized facts, suggests competition is main driver of differences in tech adoption. With PCs market-specific obsolescence is a powerful force With Apple, there is no market-specific obsolescence. For personal computers, competition increases rate of tech adoption (consistent with Arrow school of thought).
Conclusion Document stylized facts of the computer industry Stark differences in product entry, pricing, and sales across IBM-compatible and Apple platforms Vintage-capital model predicts that market structure plays a major role in explaining these differences The model predicts that competition drives faster technology adoption
Model parameters Product quality Frontier value ν 1 fixed 10 Quality growth rate γ fixed 1.029 Utility ratio ζ fixed 0.01 Income (Beta distribution) Lower bound a fixed 1 Upper bound b flexible 14.25 Density κ 1 flexible 0.715 Density κ 2 flexible 1.371 Market size ms flexible 14.575 Cost marginal cost mc flexible 0.847