A comprehensive analysis of preferences for online and offline shopping: differences across products, consumers and shopping stages

Enjoy Shop 
quick 
Large 
select 
Best 
price 
See-
touch 
Personal 
service 
Speedy 
delivery 
No hassle 
exchange 
Brand 
name 
Airline 
tickets 
5.38 
(1.85) 
5.45 
(1.69) 
5.52 
(1.67) 
5.74 
(1.61) 
3.38 
(1.80) 
3.89 
(2.08) 
4.89 
(1.87) 
4.08 
(1.91) 
4.66 
(1.66) 
Books 
3.46 
(2.13) 
4.69 
(2.00) 
4.64 
(2.09) 
4.94 
(1.89) 
2.41 
(1.74) 
3.24 
(1.95) 
3.61 
(2.12) 
3.46 
(2.00) 
4.11 
(1.72) 
Electronic 
3.61 
(1.98) 
4.39 
(1.92) 
4.56 
(1.85) 
4.60 
(1.84) 
2.18 
(1.66) 
3.11 
(1.92) 
3.50 
(1.98) 
3.25 
(1.84) 
3.96 
(1.66) 
Clothing 
2.42 
(1.93) 
3.53 
(2.21) 
3.43 
(2.14) 
3.52 
(1.97) 
1.72 
(1.47) 
2.70 
(1.82) 
2.86 
(1.96) 
2.78 
(1.85) 
3.38 
(1.74) 
Computer 
3.89 
(1.92) 
4.38 
(1.88) 
4.57 
(1.92) 
4.69 
(1.78) 
2.17 
(1.58) 
3.12 
(1.83) 
3.61 
(1.95) 
3.27 
(1.85) 
3.90 
(1.71) 
(1 = offline much better, 7 = online much better) (Standard deviations in parentheses) 
(b) University Sample 
Enjoy Shop 
quick 
Large 
select 
Best 
price 
See-
touch 
Personal 
service 
Speedy 
delivery 
No hassle 
exchange 
Brand 
name 
Airline 
tickets 
4.54 
(2.05) 
5.89 
(1.60) 
5.89 
(1.48) 
5.98 
(1.34) 
1.97 
(1.37) 
2.27 
(1.44) 
4.19 
(1.92) 
3.12 
(1.55) 
4.06 
(1.35) 
Books 
2.99 
(1.88) 
5.20 
(1.78) 
5.26 
(1.82) 
5.52 
(1.41) 
1.69 
(1.34) 
2.22 
(1.40) 
3.27 
(1.87) 
2.63 
(1.49) 
3.70 
(1.12) 
Electronic 
2.90 
(1.76) 
4.82 
(1.76) 
5.39 
(1.70) 
5.20 
(1.57) 
1.53 
(1.12) 
2.11 
(1.32) 
3.30 
(1.82) 
2.64 
(1.34) 
3.73 
(1.25) 
Clothing 
2.76 
(1.96) 
4.88 
(1.97) 
4.67 
(1.96) 
4.71 
(1.70) 
1.37 
(0.87) 
2.10 
(1.30) 
3.12 
(1.84) 
2.46 
(1.30) 
3.63 
(1.31) 
Computer 
3.60 
(1.94) 
5.01 
(1.72) 
5.28 
(1.75) 
5.28 
(1.53) 
1.71 
(1.12) 
2.22 
(1.35) 
3.30 
(1.74) 
2.79 
(1.42) 
3.79 
(1.34) 
Page 285

---

Levin et al.: Online/Offline Shopping Preferences 
Online/Offline Preference Ratings 
Table 3 gives the mean rating of online/offline shopping preference for each product at each stage (search and 
final purchase) for each sample. Ratings are on a scale of 1 (very much prefer offline) to 6 (very much prefer 
online). Consistent with H3, online preferences are greatest in each sample for airline tickets and offline preferences 
are greatest for clothing. For all products, online preference is greater at the search stage than at the purchase stage, 
thus supporting H4. The drop-off differs across products, being greatest for electronic products, but is remarkably 
similar across samples. Within each sample, analysis of variance confirmed that online/offline preferences differed 
significantly across products, 
F
(4,852) = 66.84 and 
F
(4,788) = 160.14 for the nationwide and student samples, 
respectively; and that preferences differed significantly across shopping stages, 
F
(1,213) = 103.76 and 
F
(1, 197) = 
189.37 for the respective samples. For each sample the difference in drop-off across stages for the different 
products resulted in a significant product by stage interaction, 
F
(4,852) = 9.47 and 
F
(4,788) = 11.47 for the 
respective samples, 
p
< .01 in each case.
Table 3. Mean Online/Offline Preference Ratings 
(a) Nationwide Sample 
Airline Tickets 
Books 
Electronic Clothing Computer 
Search 
5.09 (1.53) 
3.84 (1.89) 
3.83 (1.93) 
2.95 (1.85) 
4.07 (1.83) 
Purchase 
4.21 (1.88) 
3.23 (1.92) 
2.51 (1.62) 
2.37 (1.65) 
3.22 (1.80) 
(1 = much prefer offline, 6 = much prefer online) (Standard deviations in parentheses) 
(b) University Sample 
Airline Tickets 
Books 
Electronic Clothing Computer 
Search 
5.58 (0.92) 
3.68 (1.73) 
3.64 (1.83) 
2.51 (1.69) 
4.28 (1.61) 
Purchase 
4.67 (1.47) 
2.93 (1.58) 
2.26 (1.36) 
1.87 (1.21) 
3.41 (1.64) 
Explaining Online/Offline Shopping Preference Differences between Products 
The differences discussed above support the importance of product-attribute perceptions in driving 
online/offline shopping preferences for different products. We now directly examine the extent to which differences 
in online/offline shopping preferences across products can be accounted for by a weighted sum of attribute values.
The attribute values are the ratings in Table 2 of the extent to which a particular attribute for a particular product is 
thought to be better delivered online or offline. The weights correspond to the importance ratings in Table 1. For a 
given product, the corresponding values in the cells of Tables 1 and 2 are multiplied together and then the resultant 
values are summed across attributes (columns) to arrive at a weighted sum for each product. These weighted sums 
are then correlated with the product evaluations in Table 3, separately for the Search and Purchase stages of each 
product. For the Search stage, “speedy delivery” and “no hassle exchange” are excluded from the analysis because 
these represent post-search factors; for the Purchase stage, “large selection” is excluded because it pertains to search.
We thus arrive at two separate weighted sums for each product, one for the Search stage and one for the Purchase 
stage. The resulting correlations between the weighted sums and mean online/offline preference ratings are as 
follows: for the nationwide sample, 
r
= .98 and .86 for the Search and Purchase stages, respectively; for the 
University sample, 
r
= .89 and .83 for the respective stages. Thus, consistent with H3, different preferences for 
shopping online or offline for different products can be well accounted for by the extent to which attributes that are 
thought to be important for a given product are perceived to be better delivered online or offline. This was true for 
both the university and nationwide samples. 
Individual Differences 
Whereas the previous sections focus on online/offline preference differences between products averaged across 
consumers, this section focuses on differences between consumers. We examine the extent to which the variance 
observed in Table 3 can be explained for each product at each stage. The validity of these individual differences in 
online/offline shopping preference ratings is supported by significant correlations between these ratings and reported 
actual online purchasing behavior for both samples. That is, respondents who showed the greatest online preference 
in their ratings also indicated the most spent on Internet shopping.
We start by examining how demographic variables relate to differences in attribute weighting within each 
sample. For this set of analyses, the individual difference measures of age, gender, disposable income, and self-
rated computer literacy were correlated with attribute weights averaged across products. Across samples, self-rated 
computer literacy had the greatest influence on attribute weighting. For both samples, those higher in computer 
literacy placed lesser importance on the need to see-touch-handle the product, r
 
= -.26 and -.23, respectively, for the 
nationwide and student samples, p < .001 in each case. For the nationwide sample, higher computer literacy was 
Page 286

---

Journal of Electronic Commerce Research, VOL 6, NO.4, 2005 
associated with greater weight placed on enjoying the shopping experience (r = .16, p < .05). There were also 
tendencies for higher computer literacy to be associated with greater weight for large selection (r = .13, p = .056), 
speedy delivery (r = .13, p = .068), and shopping quickly (r = .12, p = .07). For the student sample, higher computer 
literacy was associated with lower importance of personal service (r = -.18, p = .01). Within the student sample, 
higher disposable income was associated with lower importance of the need to see-touch-handle the product (r = -
.14, p < .05) and greater importance placed on trusted brand name (r = .16, p < .05). Within the nationwide sample, 
age was positively related to need for speedy delivery (r = .14, p < .05). Gender was not a factor for any attribute in 
either sample.
Next, individual differences within products are examined. Table 4 summarizes the regression analyses used to 
determine the extent to which differences between consumers’ online/offline shopping preferences for each product 
at each stage can be accounted for by the various measures of individual difference. Stepwise regression analysis 
was used where the initial step (Step 1) was entering the weighted sum of attribute values for each consumer. The 
next step (Step 2) was entering a set of demographic variables consisting of age, gender, disposable income, and 
self-rated computer literacy to determine their contribution beyond individual differences in attribute weights
1
.
Individual differences in assigning values and weights to the different attributes of a product (Step 1) accounted 
for a significant proportion of the variance in online/offline preferences for each product at each stage for each 
sample. The change in 
R
2
was statistically significant (
p
< .05) or approached significance (
p
< .10) in 7 out of 10 
tests for the university sample and for 8 out of 10 tests for the nationwide sample. 
Among the demographic variables, the single best predictor of online/offline shopping preferences was self-
reported computer literacy which was a significant source of variance or approached significance for 9 out of 10 of 
the product-by-stage level regression analyses for the university sample and 5 out of 10 for the nationwide sample.
Disposable income was positively related to online shopping preference for 5 out of 10 product-by-stage 
combinations for the nationwide sample but not for the university sample where the range was restricted. Other 
factors had product-specific effects. In both samples females were more likely than males to prefer online shopping 
for clothing at both stages. Within the university sample, males were more likely to prefer online shopping for 
electronic and computer products. Within the nationwide sample, older consumers were more likely to prefer online 
search for computer products, and within the university sample, younger students were more likely to prefer online 
search for electronic products. 
Table 4A. Summary of Regression Analysis: Nationwide Sample 
Airline Tickets – Search 
Airline Tickets - Purchase 
Model 
R
2
F
change 
df p 
Model 
R
2
F
change 
df p 
1 .314 96.91 1;212 
.000 
1 .091 21.23 1;212 
.000 
2 .338 1.94 4;208 
.105 
2 .155 3.91 4;208 
.004 
Books – Search 
Books - Purchase 
Model 
R
2
F
change 
df p 
Model 
R
2
F
change 
df p 
1 .090 20.86 1;212 
.000 
1 .170 43.53 1;212 
.000 
2 .130 2.41 4;208 
.050 
2 .205 2.24 4;208 
.066 
1
The model regressed in Table 4 can be stated as follows: 
P
pi
= 
Σ
w
ji
x 
ν
ji
+ 
ε
i
j 
where P
pi
is the online-offline preference for product p by individual i; w
ji
and 
ν
ji
are the weight and value, 
respectively, assigned by individual I to attribute j; and 
ε
i
is a vector of demographic variables for individual i.
Before applying this model, we tested for normality of the criterion measures (preference ratings on a 1-6 scale). As 
can be seen in Table 3, measures such as search and purchase of airline tickets and search for computer products 
display strong online preferences, and measures such as search and purchase of clothing and purchase of electronic 
products display strong offline preferences. Not surprisingly, skewness and kurtosis were beyond acceptable limits 
for these measures. Additional analyses were performed on these measures to normalize them, using cubes or 
square roots of the rating data depending on the direction of skew. The only substantive difference from results 
reported in Table 4 was for airline search for the university sample which was the only variable with skew > 2 and 
kurtosis >7 where the influence of consumer demographics was reduced from p < .05 to p < .07 
Page 287

---

Levin et al.: Online/Offline Shopping Preferences 
Electronic Products – Search 
Electronic Products - Purchase 
Model 
R
2
F
change 
df p 
Model 
R
2
F
change 
df p 
1 .090 21.02 1;212 
.000 
1 .126 30.49 1;212 
.000 
2 .157 4.10 4;208 
.003 
2 .182 3.55 4;208 
.008 
Clothing – Search 
Clothing - Purchase 
Model 
R
2
F
change 
df p 
Model 
R
2
F
change 
df p 
1 .172 43.90 1;212 
.000 
1 .120 29.04 1;212 
.000 
2 .188 1.08 4;208 
.368 
2 .178 3.64 4;208 
.007 
Computer Products – Search 
Computer Products - Purchase 
Model 
R
2
F
change 
df p 
Model 
R
2
F 
change 
df p 
1 .152 37.93 1;212 
.000 
1 .152 37.98 1;212 
.000 
2 .224 4.83 4;208 
.001 
2 .185 2.12 4;208 
.078 
Model 1: weighted sum of product attributes, Model 2: add consumer demographic differences 
Table 4B. Summary of Regression Analyses: University Sample 
Airline Tickets – Search 
Airline Tickets - Purchase 
Model 
R
2
F
change 
df p 
Model 
R
2
F
change 
df p 
1 .020 4.00 1;196 
.047 
1 .067 14.03 1;196 
.000 
2 .039 0.95 4;192 
.435 
2 .118 2.80 4;192 
.027 
3 .073 1.35 5;187 
.246 
3 .153 1.52 5;187 
.185 
Books – Search 
Books - Purchase 
Model 
R
2
F
change 
df p 
Model 
R
2
F
change 
df p 
1 .024 4.79 1;195 
.030 
1 .021 4.09 1;195 
.045 
2 .037 0.65 4;191 
.625 
2 .056 1.80 4;191 
.131 
3 .057 0.77 5;186 
.571 
3 .072 0.63 5;186 
.680 
Electronic Products – Search 
Electronic Products - Purchase 
Model 
R
2
F
change 
df p 
Model 
R
2
F
change 
df p 
1 .043 8.86 1;196 
.003 
1 .042 8.59 1;196 
.004 
2 .124 4.45 4;192 
.002 
2 .102 3.22 4;192 
.014 
3 .143 0.83 5;187 
.529 
3 .129 1.13 5;187 
.345 
Clothing – Search 
Clothing - Purchase 
Model 
R
2
F
change 
df p 
Model 
R
2
F
change 
df p 
1 .026 5.13 1;195 
.025 
1 .038 7.79 1;196 
.006 
2 .126 5.49 4;191 
.000 
2 .098 3.18 4;192 
.015 
3 .164 1.69 5;186 
.139 
3 .137 1.69 5;187 
.138 
Computer Products – Search 
Computer Products - Purchase 
Model 
R
2
F
change 
df p 
Model 
R
2
F
change 
df p 
1 .100 21.70 1;196 
.000 
1 .086 18.35 1;196 
.000 
2 .159 3.39 4;192 
.011 
2 .165 4.54 4;192 
.002 
3 .171 0.53 5;187 
.752 
3 .l86 0.97 5;187 
.438 
Model 1: weighted sum of product attributes, Model 2: add consumer demographic differences 
Model 3: add consumer personality differences