The Demand Effects of Product
Recommendation Networks: An Empirical Analysis of
Network Diversity and Stability
Zhijie LIN, Khim Yong GOH, Cheng Suang HENG
2 / 30
Introduction
• Example
3 / 30
Introduction
• Product Network
I
F
I I
I
F
Incoming Network Product
Focal Product
Incoming Tie
OO O
O Outgoing Network Product
Outgoing Tie
4 / 30
Introduction
• Research Gap 1: Network Bidirectionality
5 / 30
Introduction
• Research Gap 2: Network-Based Product Marketing
– which is to be recommended (network diversity)
– how often to replace recommendations (network stability)
6 / 30
Introduction
• Research Gap 3: Co-View vs. Co-Purchase
7 / 30
Introduction
• Research Questions
– Is the demand of a product influenced by both the incoming
network and outgoing network?
– How is the demand of a product influenced by product
network attributes in terms of network diversity and
network stability?
– How do the network diversity and stability effects differ
between co-view and co-purchase recommendations?
8 / 30
Literature
• Recommendation Systems
– algorithm design
• content-based approach (Adomavicius & Tuzhilin 2005)
• collaborative filtering approach (Ricci et al. 2011)
– economic impact
• decrease sales diversity (Mooney & Roy 2000)
• increase sales diversity (Brynjolfsson et al. 2006)
• increase individual consumer level diversity but decrease aggregate
market level diversity (Fleder & Hosanagar 2009)
• direct & indirect effect of recommendation (Pathak et al. 2010)
9 / 30
Literature
• Network Analysis– product network
• World Wide Web sites (Katona & Sarvary 2008)
• news reports (Dellarocas et al. 2010)
• blogs (Mayzlin & Yoganarasimhan 2012)
• videos (Goldenberg et al. 2012)
• demand shocks from exogenous book reviews (Carmi et al. 2010)
• product value = intrinsic value + extrinsic value (Carmi et al. 2011)
• network connection and demand (Oestreicher-Singer & Sundararajan 2012a)
• network connection and long tail (Oestreicher-Singer & Sundararajan 2012b)
– social network• diversity
– knowledge sharing (Cummings 2004), health condition (Barefoot et al. 2005), work performance (Wu 2011), online content propagation (Yoganarasimhan 2012)
• stability– mobile Internet usage (Ghose et al. 2011), product adoption (Tucker 2011)
10 / 30
Hypothesis
• Network Diversity– H1A: incoming network diversity � product demand (+)
• consumer heterogeneous product preference (Allenby et al. 1998)
– H1B: outgoing network diversity � product demand (-)• information diversity (Berlyne 1960) and attention distraction (Geissler et al. 2006)
• Network Stability– H2A: incoming network stability � product demand (+)
• visits-to-purchase (Moe & Fader 2004)
• consumer habit in online searches (Moe & Yang 2009)
– H2B: outgoing network stability � product demand (-)• substitutes: more desirable products (Schafer et al. 2001)
• complements: options for co-purchase (Schafer et al. 2001)
• Co-view vs. Co-purchase– H3: co-purchase > co-view
• co-purchase information is more persuasive
• more strategic use of co-purchase recommendations (e.g., Amazon)
11 / 30
Empirical Analysis
• Data Source
– Tmall.com
• Chinese-language business-to-consumer (B2C) platform
• the most visited B2C website in China (Alexa 2012)
• Nikon exclusive store (camera, lens, battery, flash, accessory)
– daily snapshots of product-related information at 12:00 a.m.
– daily snapshots of product network structure at 12:00 a.m.
– detailed product transaction records
• recommendation algorithm
– collaborative filtering approach
12 / 30
Empirical Analysis
13 / 30
Empirical Analysis
(buyer) (product name) (product price) (quantity) (transaction date & time)
(previous) (next)
14 / 30
Empirical Analysis
(product name)
(product price)
(product review)
(Co-view)
(Co-purchase)
(product name)
(product price)
(product review)
15 / 30
Empirical Analysis
• Focal Variables
– sales quantity
• QUANit
• total sales quantity of product i on day t
– network diversity
• co-view: ID_CVit / OD_CVit• co-purchase: ID_CPit / OD_CPit• number of unique product categories in product i’s incoming / outgoing
network on day t
– network stability
• co-view: IS_CVit / OS_CVit• co-purchase: IS_CPit / OS_CPit• percentage of overlap of network connections in product i’s incoming /
outgoing network of day t and day t+1
16 / 30
Empirical Analysis
• Control Variables
– product level
• price, review volume, review rating, monthly sales quantity, inventory, bookmark frequency
– product category level
• category fixed effect
– product network level
• in-degree and out-degree centrality
• Incoming and outgoing network product price
• Incoming and outgoing network product review volume
• Incoming and outgoing network product review rating
• Incoming and outgoing network product demand
– time unit level
• time fixed effect (daily)
17 / 30
Empirical Analysis
• Implicit Demand Correlation
SubstituteComplement
18 / 30
Empirical Analysis
• Implicit Demand Correlation
– cross-category price elasticities (Manchanda et al. 1999)
– category
• accessory, battery, flash, lens, compact camera, single lens reflex
(SLR) camera
– model
( )6 6 6
1 1 1
6 6 6
1 1 1
ln mn mn m
n n
n
mn mn mn
n
mt
n n n
n n n
n n n
nt nt nt
nt nt n mt mt
QUAN LP PV PR
PS IN BM
α β χ
δ ε φ ω ξ
= = =
= = =
= = =
= = =
= + +
+ + + + +
∑ ∑ ∑
∑ ∑ ∑
19 / 30
Empirical Analysis
• Implicit Demand Correlation
– price elasticity matrix
Category m
Category n
(1)
Accessory
(2)
Battery
(3)
Compact
Camera
(4)
Flash
(5)
Lens
(6)
SLR
Camera
(1) Accessory -0.127***
(0.047)
-0.089**
(0.041)
-0.055***
(0.008)
0.043***
(0.008)
0.071***
(0.016)
0.006
(0.025)
(2) Battery 0.245***
(0.076)
-0.163*
(0.094)
0.032*
(0.018)
0.005
(0.012)
0.151***
(0.012)
0.121*
(0.073)
(3) Compact Camera -0.209
(0.318)
-0.183***
(0.070)
-0.469***
(0.135)
0.050**
(0.023)
-0.043
(0.030)
-0.034
(0.024)
(4) Flash -0.796*
(0.458)
-0.323**
(0.153)
-0.338***
(0.092)
-0.703***
(0.077)
-0.015
(0.148)
-0.144
(0.125)
(5) Lens 0.183***
(0.024)
-0.012
(0.051)
0.043**
(0.021)
-0.007
(0.007)
-0.832***
(0.060)
0.013
(0.079)
(6) SLR Camera 0.293
(0.325)
0.240
(0.173)
-0.080
(0.145)
-0.022
(0.021)
0.094
(0.155)
-0.881***
(0.267)
Note: *p<0.1, **p<0.05, ***p<0.01
Note: Entry (m, n) represents the elasticity of category m‘s demand with respect to the price of category n.
20 / 30
Empirical Analysis
• Implicit Demand Correlation– Eij: the elasticity of product i’s demand with respect to the price of
product j
– i in category m, j in category n, m ≠ n� Eij = entry(m, n)
– i in category m, j in category n, m = n� Eij = |entry(m, n)|
• assumption: products within a category are substitutes (Walters 1991)
– for focal product i
• product j is a substitute if Eij > 0
• product j is a complement if Eij < 0
– substitution & complementarity
1
* , 0=
=
>=∑s S
ii s st i
s
t sE QUAN for all s with ESE
1
* , 0=
=
<=∑c C
ii c ct i
c
t cE QUAN for all c with ECE
21 / 30
Empirical Analysis
• Model Specification
( ) 1 2 3 4
5 76 8
9 10
15 16 1
11 12
7 1
13 1
1 0
8
4
9 2_
l
_
n _ _ _ _
_ _ _ _
_ _ _ _
it it it it it
it it it it
it it it it it it
it it it it
it
QUAN ID CV ID CP OD CV OD CP
IS CV IS CP OS CV OS CP
LP PV PR PS IN BM
IC CV IC CP OC CV OC CP
ILP CV ILP C
β β β β
β β β β
β β β β β β
β β β β
β β
= + + +
+ + + +
+ + + + + +
+ + + +
+ + 21 22
23 24 25 26
27 28 29 30
31 32 33 34
35 36
_ _
_ _ _ _
_ _ _ _
_ _ _ _
i t
it it it
it it it it
it it it it
it it it it
it it i it
P OLP CV OLP CP
IPV CV IPV CP OPV CV OPV CP
IPR CV IPR CP OPR CV OPR CP
IQ CV IQ CP OQ CV OQ CP
SE CE C T
β β
β β β β
β β β β
β β β β
β β γ µ α ε
+ +
+ + + +
+ + + +
+ + + +
+ + + + + +
22 / 30
Empirical Analysis
• ResultNote: *p<0.1, **p<0.05, ***p<0.01
VariableRE
Control
Price -0.000
Review Volume 0.002***
Review Rating 0.004***
Monthly Sales 0.012***
Inventory 0.000***
Number of Bookmarks -0.000***
In-degree Centrality (Co-view) -0.000
In-degree Centrality (Co-purchase) -0.002***
Out-degree Centrality (Co-view) -0.011***
Out-degree Centrality (Co-purchase) 0.001
Incoming, Price (Co-view) -0.000
Incoming, Price (Co-purchase) 0.000
Outgoing, Price (Co-view) 0.000
Outgoing, Price (Co-purchase) 0.000
VariableRE
Control
Incoming, Review Volume (Co-view) 0.000
Incoming, Review Volume (Co-purchase) -0.000
Outgoing, Review Volume (Co-view) 0.000
Outgoing, Review Volume (Co-purchase) -0.000
Incoming, Review Rating (Co-view) -0.002**
Incoming, Review Rating (Co-purchase) 0.000
Outgoing, Review Rating (Co-view) 0.002*
Outgoing, Review Rating (Co-purchase) -0.001
Incoming, Sales Quantity (Co-view) 0.002
Incoming, Sales Quantity (Co-purchase) 0.005***
Outgoing, Sales Quantity (Co-view) 0.004**
Outgoing, Sales Quantity (Co-purchase) 0.004***
Substitution Effect -0.005***
Complementarity Effect 0.009***
23 / 30
Empirical Analysis
• ResultVariable (1)
RE
Incoming
(2)
RE
Outgoing
(3)
RE
Co-view
(4)
RE
Co-purchase
(5)
RE
Full
(6)
FE
Full
Incoming, Diversity (Co-view) -0.004 -0.005* 0.001 0.005
Incoming, Diversity (Co-purchase) 0.009*** 0.010*** 0.012*** 0.002
Outgoing, Diversity (Co-view) -0.003 -0.001 -0.007 -0.002
Outgoing, Diversity (Co-purchase) -0.005*** -0.007*** -0.007*** -0.002
Incoming, Stability (Co-view) 0.007** 0.006* 0.006 0.003
Incoming, Stability (Co-purchase) -0.002 0.000 -0.002 -0.004
Outgoing, Stability (Co-view) 0.007* 0.000 0.005 0.007*
Outgoing, Stability (Co-purchase) -0.017*** -0.015*** -0.017*** -0.012***
Substitution Effect -0.005*** -0.005*** -0.005*** -0.005*** -0.005*** -0.004***
Complementarity Effect 0.008*** 0.009*** 0.009*** 0.008*** 0.009*** 0.007***
Observations 41,379 41,379 41,379 41,379 41,379 41,379
Hausman Test χ2 = 0.00, p = 1.00
R2 0.4919 0.4916 0.4913 0.4922 0.4936 0.1017
Note: *p<0.1, **p<0.05, ***p<0.01
24 / 30
Empirical Analysis
• Simultaneity
Variable (1)
Preferred
(2)
Lag 1
(3)
Lag 7
(4)
Lead 1
(5)
Lead 7
(6) Reduce
Bi-
directionality
(=100%)
(7) Reduce
Bi-
directionality
(>50%)
(8) Reduce
Bi-
directionality
(>10%)
Incoming, Diversity (Co-view) 0.001 -0.002 -0.010*** 0.007** 0.008*** 0.002 0.005 0.009
Incoming, Diversity (Co-purchase) 0.012*** 0.008*** 0.010*** 0.012*** 0.015*** 0.012*** 0.015*** 0.021**
Outgoing, Diversity (Co-view) -0.007 -0.001 -0.005 -0.005 -0.011*** -0.006 -0.007 0.008
Outgoing, Diversity (Co-purchase) -0.007*** -0.003** -0.004*** -0.004*** -0.004*** -0.009*** -0.013*** -0.033***
Incoming, Stability (Co-view) 0.006 -0.004 0.001 0.001 0.001 0.006 0.010* 0.008
Incoming, Stability (Co-purchase) -0.002 0.000 0.005 0.002 0.005 -0.002 -0.002 -0.006
Outgoing, Stability (Co-view) 0.005 0.009** 0.004 0.002 0.005 0.006 0.011* 0.019
Outgoing, Stability (Co-purchase) -0.017*** -0.011*** -0.009*** -0.012*** -0.012*** -0.018*** -0.027*** -0.060***
Observations 41,379 40,836 38,927 40,836 38,927 34,070 22,279 8,200
Sample Reduced 17.664% 46.159% 80.183%
R2 0.4936 0.4949 0.5021 0.4943 0.4939 0.5025 0.5339 0.6268
Note: *p<0.1, **p<0.05, ***p<0.01
25 / 30
Empirical Analysis
• Elasticity
• Standardized Coefficient
Network diversity ID_CV ID_CP OD_CV OD_CP
- 0.014*** - -0.011***
Network stability IS_CV IS_CP OS_CV OS_CP
- - - -0.012***
Note: *p<0.1, **p<0.05, ***p<0.01
Network diversity ID_CV ID_CP OD_CV OD_CP
- 0.055*** - -0.031***
Network stability IS_CV IS_CP OS_CV OS_CP
- - - -0.023***
Note: *p<0.1, **p<0.05, ***p<0.01
26 / 30
Empirical Analysis
• Hypothesis Testing Results
Hypothesis Support
H1A Network diversity (incoming) � Demand (+) Yes (co-purchase)
H1B Network diversity (outgoing) � Demand (-) Yes (co-purchase)
H2A Network stability (incoming) � Demand (+) No
H2B Network stability (outgoing) � Demand (-) Yes (co-purchase)
H3 Co-purchase > Co-view Yes
27 / 30
Empirical Analysis
• Robustness (1)
Variable (1)
Preferred
(2)
Significant
Elasticity
Only
(3)
Without
Elasticity
Weight
(4)
1st-stage
Uncertainty
Incoming, Diversity (Co-view) 0.001 0.001 0.006*** 0.001
Incoming, Diversity (Co-purchase) 0.012*** 0.012*** 0.010*** 0.012***
Outgoing, Diversity (Co-view) -0.007 -0.007* -0.004 -0.006
Outgoing, Diversity (Co-purchase) -0.007*** -0.007*** -0.005*** -0.007***
Incoming, Stability (Co-view) 0.006 0.006 0.004 0.006
Incoming, Stability (Co-purchase) -0.002 -0.002 -0.000 -0.001
Outgoing, Stability (Co-view) 0.005 0.005 0.006* 0.005
Outgoing, Stability (Co-purchase) -0.017*** -0.017*** -0.012*** -0.017***
Observations 41,379 41,379 41,379 41,379
R2 0.4936 0.4926 0.6623 0.4890
Note: *p<0.1, **p<0.05, ***p<0.01
28 / 30
Empirical Analysis
• Robustness (2)
Variable (1)
Preferred
(2)
Mean
Centering
(3)
Standardi
-zation
(4)
Hetero-
skedasticity
(5)
Serial
Correlation
(6)
Price
Endogeneity
(7)
Relative
Difference
Incoming, Diversity (Co-view) 0.001 0.001 0.001 0.001 0.003 0.002 0.001
Incoming, Diversity (Co-purchase) 0.012*** 0.012*** 0.015*** 0.012* 0.007*** 0.014*** 0.012***
Outgoing, Diversity (Co-view) -0.007 -0.007 -0.003 -0.007* -0.004 -0.010* -0.007
Outgoing, Diversity (Co-purchase) -0.007*** -0.007*** -0.008*** -0.007 -0.004** -0.008*** -0.007***
Incoming, Stability (Co-view) 0.006 0.006 0.002 0.006 0.005 0.008 0.006
Incoming, Stability (Co-purchase) -0.002 -0.002 -0.001 -0.002 -0.003 -0.002 -0.002
Outgoing, Stability (Co-view) 0.005 0.005 0.002 0.005* 0.006 0.005 0.005
Outgoing, Stability (Co-purchase) -0.017*** -0.017*** -0.006*** -0.017* -0.013*** -0.020*** -0.017***
Observations 41,379 41,379 41,379 41,379 41,379 29,172 41,379
R2 0.4936 0.4936 0.4936 0.4936 0.4754 0.5037 0.4936
Note: *p<0.1, **p<0.05, ***p<0.01
29 / 30
Empirical Analysis
• Robustness (3)
Variable (1)
Preferred
(2)
Price
Diversity
(3)
Market
Share
(4)
Actual
Day
(5)
Weekly
(6)
Dell
PC
Incoming, Diversity (Co-view) 0.001 0.000 -0.007 -0.001 0.047*** 0.001
Incoming, Diversity (Co-purchase) 0.012*** 0.000** 0.641*** 0.008*** 0.071*** 0.006**
Outgoing, Diversity (Co-view) -0.007 -0.000 0.034 -0.003 -0.063*** -0.003
Outgoing, Diversity (Co-purchase) -0.007*** -0.000** -0.274*** -0.005*** -0.032*** -0.016***
Incoming, Stability (Co-view) 0.006 0.004 0.055 0.003 0.039* 0.010
Incoming, Stability (Co-purchase) -0.002 -0.005 -0.004 -0.001 0.028 0.010
Outgoing, Stability (Co-view) 0.005 0.008** 0.361** 0.007* 0.050** 0.028*
Outgoing, Stability (Co-purchase) -0.017*** -0.015*** -0.451*** -0.019*** -0.102*** -0.027***
Observations 41,379 41,379 41,379 41,379 7,800 15,616
R2 0.4936 0.4929 0.4037 0.4938 0.6508 0.4597
Note: *p<0.1, **p<0.05, ***p<0.01
30 / 30
Contribution
• Theoretical
– product network / product recommendation
• unidirectional vs. bidirectional
• positive vs. negative impact of recommendation
• diversity effect and stability effects
• co-view vs. co-purchase
• Practical
– retailers: marketing strategies
• diversity: high incoming (co-purchase) + low outgoing (co-purchase)
• stability: low outgoing (co-purchase)
• more saliently display co-purchase recommendations
– platform providers: recommendation systems design
Q & A
Thank You
The Demand Effects of Product
Recommendation Networks: An Empirical Analysis of
Network Diversity and Stability
32 / 30
Overview
• Introduction
• Literature
• Hypothesis
• Empirical Analysis
• Contribution
• Conclusion
33 / 30
Empirical Analysis
• Finding
Co-view
Co-purchase
Incoming Outgoing
Diversity (+)Diversity (-)
Stability (-)
^
34 / 30
Conclusion
• Limitations
– unable to control for all potential sources of endogeneity
– lack of individual consumer clickstream data
• Future Research
– spillover effects of WOM (e.g., consumer review)
– automated recommendations vs. WOM recommendations
35 / 30
Appendix
• Descriptive Statistics
Variable Mean Std. Dev. Min Max
Sales Quantity 2.283 6.791 0.000 680.000
Incoming, Diversity (Co-view) 0.888 0.584 0.000 6.000
Incoming, Diversity (Co-purchase) 1.180 1.250 0.000 6.000
Outgoing, Diversity (Co-view) 1.090 0.446 0.000 4.000
Outgoing, Diversity (Co-purchase) 1.519 1.133 0.000 5.000
Incoming, Stability (Co-view) 0.787 0.307 0.000 1.000
Incoming, Stability (Co-purchase) 0.711 0.380 0.000 1.000
Outgoing, Stability (Co-view) 0.757 0.321 0.000 1.000
Outgoing, Stability (Co-purchase) 0.727 0.378 0.000 1.000
Note: Observations = 41,379; Products = 257; Days = 184.
36 / 30
Appendix
• Correlation Matrix
Variable 1 2 3 4 5 6 7 8 9
1 Sales Quantity -
2 Incoming, Diversity (Co-view) -0.027 -
3 Incoming, Diversity (Co-purchase) 0.000 0.478 -
4 Outgoing, Diversity (Co-view) -0.068 0.212 0.030 -
5 Outgoing, Diversity (Co-purchase) -0.024 0.201 0.472 0.294 -
6 Incoming, Stability (Co-view) 0.020 -0.184 -0.081 0.060 0.066 -
7 Incoming, Stability (Co-purchase) 0.011 -0.143 -0.247 0.035 -0.025 0.146 -
8 Outgoing, Stability (Co-view) 0.013 0.030 0.032 -0.242 -0.009 0.193 0.100 -
9 Outgoing, Stability (Co-purchase) -0.000 -0.014 -0.097 -0.226 -0.303 0.026 0.112 0.434 -
Note: Only major variables are reported. The correlations of these variables with other variables are generally small.
37 / 30
Appendix
• Granger Causality Test
No. of
lags
ID_CV ID_CP OD_CV OD_CP IS_CV IS_CP OS_CV OS_CP
χ2 p χ2 p χ2 p χ2 p χ2 p χ2 p χ2 p χ2 p
1 0.638 0.425 0.024 0.876 2.585 0.108 0.511 0.475 2.359 0.125 2.731 0.098 2.404 0.121 3.054 0.081
2 2.902 0.234 0.653 0.722 3.159 0.206 0.570 0.752 3.130 0.209 3.814 0.148 3.335 0.189 4.997 0.082
3 1.934 0.586 0.977 0.807 1.658 0.646 1.392 0.707 4.759 0.190 4.980 0.173 4.240 0.237 6.169 0.104
4 2.253 0.689 2.715 0.607 3.913 0.418 4.265 0.371 8.189 0.085 7.910 0.095 4.472 0.346 6.735 0.151
5 3.626 0.604 4.631 0.463 4.154 0.528 4.011 0.548 7.954 0.159 8.630 0.125 4.349 0.500 7.496 0.186
6 3.623 0.728 5.855 0.440 3.511 0.742 4.183 0.652 7.534 0.274 8.785 0.186 5.460 0.486 8.700 0.191
7 3.700 0.814 7.405 0.388 3.633 0.821 4.300 0.745 8.251 0.311 10.569 0.159 5.640 0.582 9.051 0.249