www.ccsenet.org/ibr International Business Research Vol. 3, No.

4; October 2010

Determinants of Recent Online Purchasing and the Percentage

of Income Spent Online
Brendan Hannah & Kristina M. Lybecker
Department of Economics and Business, Colorado College
14 E. Cache La Poudre, Colorado Springs, Colorado, United States 80903
E-mail: hannah.brendan@gmail.com / kristina.lybecker@coloradocollege.edu
Abstract
The recent stagnation of electronic commerce highlights the need to understand contemporary online consumer
behavior. E-commerce’s slow growth has coincided with an explosion in the usage of Web 2.0 activities. These
novel online venues have created many new channels for online retailers to reach buyers, yet these online
activities have gone largely unstudied. This study incorporates current user demographics and Web 2.0 activities
to dynamically model the determinants of two key measurements of recent online shopping, a recent purchase
and the novel dependent variable, percentage of income spent online. Regression analysis is applied to a
nationally representative 2007 survey of the U.S. online population. Determinants of a recent online purchase
include, ownership of a credit card, PayPalTM account, listening to podcasts, participating in online auctions, and
for the first time, female gender. In a second regression, positive determinants for the percentage of income spent
online include male gender, educational attainment, online auctions, instant messaging, and online dating. Online
spending increases with time online, and appears to compete with other forms of online entertainment and social
networking. These results produce snapshots of contemporary online shoppers that can be used by electronic
retailers to determine which product characteristics to highlight for greatest impact, and to efficiently target
specific Web 2.0 activities, such as entertainment, podcast and social network websites, to develop new and
robust marketing platforms.
Keywords: Online consumer behavior, Online-shopping, E-commerce, Web 2.0 marketing
1. Introduction
The number of Internet users in the U.S. has almost doubled over the last eight years to approximately 220
million in 2008 (Nielsen 2008) with over 70% purchasing online (Pew Internet 2009a). This expansion in the
Internet population has coincided with an explosive increase in the online usage of Web 2.0 activities such as
social networking, video sharing, and other entertainment sites. These novel online venues have created many
more channels for online retailers to reach buyers. Yet yearly growth of e-commerce has fallen short of virtually
all analysts’ predictions for nearly a decade (Bakos 2001). The percentage of online consumers has remained
steady at about 70% since late 2004 (Pew Internet, 2008b) when online buying represented 2.5% of national
retail sales (U.S. Census Bureau, 2008). E-commerce has only grown about 0.6% in relation to retail sales over
the last four years with revenue just over $144 billion in 2008 (U.S. Census Bureau, 2009; 2010).
Web 2.0 channels present online retailers with novel opportunities to efficiently target potential buyers and
market their goods and services. Two noteworthy trends indicate that the success of electronic retailers will
depend on cultivating existing consumers: first, the proportion of consumers who never buy online remains
steady, and second, over 85% of online shoppers continue to buy each year (Lohse et al., 1999; Pew Internet,
2008b). The ability of e-retailers to grow their markets is contingent on a thorough understanding of
contemporary online behavior. Most published studies focus on the decision to adopt Internet shopping rather
than recent purchasing behavior and utilize data from the 1990’s, a period in which online buyers were
predominantly older, wealthier, well-educated males (Swinyard and Smith, 2003). However, current user
demographics have shifted to equal percentages of men and women buying online (Pew Internet, 2008a;b) who
are younger and possess lower levels of income and education (Cummings and Kraut, 2002). This study seeks to
fill this gap by analyzing more current online purchasing behavior and suggesting how online retailers can locate
and appeal to potentially receptive buyers.
The contributions of this paper are twofold. First, by utilizing a more recent dataset and including Web 2.0
activities, we are able to unearth the direct links between Internet users and their online shopping decisions.
Second, by examining the amount of income spent online, we can propose real world applications for these
results and discuss marketing implications for reaching buyers through Web 2.0 channels. To accomplish this,
we review previous literature on the determinants of online shopping and current online activities patronized by
these consumers. Next, we explain how these variables provide the methodological foundation for the models

60 ISSN 1913-9004 E-ISSN 1913-9012

www.ccsenet.org/ibr International Business Research Vol. 3, No. 4; October 2010

and the implementation of these models using this dataset. After presenting the findings and the limitations of
the analysis, we apply the findings and suggest paths for future research.
2. Existing Literature
Online shopping has provided researchers with a rich set of questions and wealth of new information on
consumer behavior and decision-making. Dating from the late 1990s, these studies provide both scholars and
retailers with additional perspective on how the Internet has altered consumption and the factors that have
facilitating this change. Information on the demographics, socioeconomics and consumer behavior of Internet
shoppers has recently been synthesized into an Online Shopping Acceptance Model (OSAM) by Zhou et al.
(2007). Online purchasing appears to be most related to convenience (Zhou et al., 2007) in addition to recreation
and economic advantages (Donthu and Garcia, 1999; Korgaonkar and Wolin, 1999; Li et al. 1999; Swaminathan
et al., 1999). The higher efficiency of e-commerce has reduced buyer search costs (Bakos, 1997) and produced
lower prices for several online goods and services than their offline counterparts (Brown and Goolsbee, 2002;
Brynjolfsson et al., 2003; Lee et al., 2003), offering the promise of products supplying good value to economic
shoppers.
Online purchasing has also been facilitated by enhanced Internet accessibility as prices have steadily dropped
and connections have become faster. Through this, online usage has developed into a daily part of nearly all
American’s lives for email correspondence and to obtain information (Nielsen 2008). In 2008, just over 90% of
American home Internet consumers connected using broadband (WebSiteOptimization.com 2008). Higher
connection speeds allow the Internet to be used more heavily for Web 2.0 activities, placing online gaming,
instant messaging and social networking as the most time-intensive activities among broadband consumers
(Nielsen//NetRatings, 2006). Between 32% and 35% of users visit blogs and social network websites such as
MySpace, Facebook and others (Pew Internet 2009a). The same study showed that 28% access or download
digital content, 26% participate in eBay and other online auctions, 52% watch video sites including YouTube
and its derivatives, and 35% play games online (Pew Internet, 2008b). Advertising through these Web 2.0
channels allows online retailers to capitalize on the huge popularity of these novel online venues. Patronization
of these activities can provide online retailers with detailed information on potential buyers’ interests and allows
for highly efficient marketing at lower costs but remains an unstudied aspect of online purchasing behavior.
To our knowledge, this study is the first in nearly a decade to define the factors that influence contemporary
online shopping behavior. Using representative U.S. online consumer data and the current range of Internet
activities, two models are developed to explain consumer determinants of online purchasing within the last year
and the percentage of income spent online in the last three months, a variable that has never before been studied.
There are several elements of this study that both set it apart as a unique contribution and mark it as a valuable
addition to the existing literature. First, this study delineates factors that influence online spending rather than the
decision to adopt or continue online shopping. The empirical work also utilizes a comprehensive dataset
reflecting the newly-diverse, current demographics and online activities of online shoppers, information that can
provide insight for online retailers. Since a large portion of the online population uses the Internet primarily for
entertainment purposes and thus has greatly increased usage (Pew Internet, 2008b; 2009a), this study is the first
to investigate the relationships between utilization of these new activities and online purchasing. The integration
of demographic and socioeconomic variables with Web 2.0 activities provides the basis for developing the
models presented here that describe the purchasing decisions of the current online population. These findings
provide tangible evidence that can be utilized in marketing decisions, highlighting which gender-specific
strategies will likely be the most successful. Finally, the study suggests that the greatest promise of online
retailing, lower prices through reduced search costs, may not actually be its greatest attraction for today’s
consumers.
3. Methodology
3.1. The Models
Model 1. The first model of online purchasing within the last year (purchase, y1) incorporates five categories of
variables (demographic, socioeconomic, Internet usage, product perceptions and alternative activities). The
dependent variable, y1, is a Bernoulli random variable: it has only two outcomes, purchase within the last year (1)
or no purchase (0). A linear regression model does not fit a binary dependent variable such as y1, so a logistic
regression model is used. The logistic model maps a linear combination of the predictors to the probability that
the associated dependent variable equals 1 (Menard, 2002; Gelman and Hill, 2006). Y1 is analyzed using a
weighted (Note 1) logistic regression to ensure the results are nationally representative (Note 2). Table 1
provides a description and statistics of each variable.
y1  Bernoulli (p)

Published by Canadian Center of Science and Education 61

www.ccsenet.org/ibr International Business Research Vol. 3, No. 4; October 2010

1
(Equation 1) p=

26
(  0   1 1 )
1 e i 1

Model 2. The second model examines the percentage of income spent online within the last three months
(spending as a percentage of income, y2) and incorporates the same five categories of variables described above.
A linear regression model is applied to y2 because its range is the real line (Gelman and Hill 2006). Equation 2 is
analyzed as a weighted least squares regression.
26
(Equation 2) y2  N(ß0 + 
i1
ßixi, 2)

3.2. Data Set and Sources

The regression models utilize a data set acquired through a leading, publicly traded market research company
that specializes in e-commerce and online demographics (Note 3). Data are compiled from a portion of a
forty-two question, closed-end survey conducted in June 2007. The questions focus on demographics, product
preferences, and online behaviors, attitudes and activities. The 3,580 participants were selected by Ipsos from
their U.S. online consumer panel, and the sample was balanced by demographic and behavioral characteristics
derived from contemporaneous U.S. Census Bureau and research pertaining to the U.S. online population.. This
dataset is particularly well-suited to this study due to the retail nature of the data collected. This data is
proprietary market research and provides an accurate sample of the current online population. The survey
questions were constructed to glean information on the online user’s decision to purchase a good or service as
well as amount spent online, information tailored to the needs of online businesses.
Several shortcomings of the data set may slightly reduce the explanatory power for each of the models. The data
set presents the amount spent online and income as bracketed variables. The median amount spent online was
divided by median quarterly income to determine the percentage of income spent online during the last three
months. However, the use of median values from bracketed variables for age, amount spent online and income
does not assure a constant standard error and can lead to lower explained variance. Ceilings on income
($100,000+) and on the amount spent online ($5,000+) also reduce the accuracy of the analysis, and the lack of
specific information on high earning and spending outliers is expected to limit conclusions drawn here. The
questionnaire also lacked key variables regarding demographics (ethnic background or location),
socioeconomics (specific occupation, home ownership or number of dependents), Internet usage (user’s
technical aptitude or online fluency) and other intangible variables (perceptions of online buying or security in
online transactions,) that would be expected to further influence online consumer behavior.
4. Results and Implications for e-Commerce
The models described here utilize representative U.S. data on 2007 online usage, patronization of Internet
activities and user demographics.
4.1. Model 1
The logistic regression used for Model 1 is highly significant (p < 0.001) (Table 2). Table 4 provides details on
the 26 variables studied, of which 12 are significant. The R-squared value of 0.237 is obtained by replicating the
model for each participant and creating a predicted value (Note 4). Then a weighted, binary regression is run
with the actual participant’s outcome (purchase = 1 or no purchase = 0) as the dependent variable and the
predicted values as the independent variable. While close to 24% of the variance is explained by the 26
independent variables studied, demographic variables alone explained only 3.3% of the variance (see Table 3).
The only significant demographic determinant of a recent online purchase is the gender of the participant
(p<0.01 Table 4). To our knowledge, this is the first time female gender is shown to be positively correlated with
online shopping; in this case, purchasing within the last year. For an online user with low technological
acceptance (Note 5), when adjusted for all other variables, females are 6.6% (37.9% for women vs. 31.3% for
men are the probabilities of making on an online purchase) more likely than males to have made a recent online
purchase. In contrast, for online users of high technological acceptance, females are only 0.4% more likely to
have made a recent online purchase (98.9% vs. 98.5%) (Note 6). These result align with the shifting online
demographics as noted by Pew Internet (2008a,b); an increasing percentage of women online and becoming
more sophisticated users. Not surprisingly, this translates into a higher rate of recent online purchasing by
women than men at the time of this survey. These results provide a reason for retailers to target advertising at
websites heavily patronized by women.
The majority of socioeconomic variables showed positive correlations with a recent online purchase, providing
verification of previous studies (Bellman et al., 1999; Donthu and Garcia, 1999; Korgaonkar and Wolin, 1999;
Li et al., 1999). Even though the regression coefficient is small, household income is positively correlated with
recent online buying. For a male user with low technological acceptance, an income increase from $30,000 to

62 ISSN 1913-9004 E-ISSN 1913-9012

www.ccsenet.org/ibr International Business Research Vol. 3, No. 4; October 2010

$100,000, translates to a 15.1% (46.4% vs. 31.3%) higher probably of having purchased an item online within
the last year. Male online users with low technological acceptance with a payment account such as PayPal™ are
15.0% (46.3% vs. 31.3%) more likely to have shopped within the last year. Female users with low technological
acceptance have 23.4% (37.9% vs. 14.5%) higher likelihood of purchasing an item online if they possess a credit
card. The results show that marketing preferentially to high earners and increasing access to credit cards or
PayPal™ accounts should increase the adoption of online shopping.
The Internet usage determinants include several previously studied variables on communication and information
retrieval. In contrast to the findings of previous studies (Huang, 1998; Bellman et al., 1999; Bhatnagar et al.,
2000), weekly Internet usage is not significantly correlated with a recent purchase. Use of email, online research
of products, use of a search engine, software downloads, and investigation of travel online is all positive,
significant determinants of recent online shopping. Use of Instant Messenger for communication, a previously
unstudied variable, is not significantly correlated.
Several Web 2.0 variables describing contemporary Internet activities such as shopping behavior, entertainment
and social networking are also included in the regression models. Of these, participation in online auctions is the
most significant determinant, followed by listening to podcasts. The remaining five variables are not significant
determinants of recent online purchasing (blogs, online gaming, online dating, social networking, classified ads).
These findings demonstrate that many Web 2.0 channels, such as social networking sites are not effective
advertising channels to induce an online user’s first purchase.
These data include measures of product perceptions in order to account for their influence on contemporary
online consumer behavior. Recent online purchasing is strongly correlated with a user preference for products
that are described as new and innovative while the variable for products described as good values is statistically
insignificant. These findings indicate that new, dynamic product lines should highlight innovativeness in their
descriptions and advertising.
Contrary to Bellman et al.’s (1999) theory of time starvation, time spent working proved insignificant in
explaining recent online purchasing. In addition, given that weekly television hours is not a determinant,
electronic retailers should be less motivated to advertise on television, at least to heavier television watchers.
Their resources might be better spent advertising via new channels on the Internet that allow for targeted
marketing, such as podcasts. Most users apparently view Internet shopping as a time saving activity, driven by
convenience and the ability to find novel and hard to find items.
4.2. Model 2
Model 2 Using All Data. The second model utilizes a novel dependent variable, the percentage of income spent
online in the last three months. As reported in Table 2, the regression is highly significant. Table 4 provides the
coefficients and describes the significance of the 26 determinants. These variables explain 19.4% of the variance,
while demographic variables account for only 2.1% (see Table 3).
The second model reveals that male gender and higher educational attainment are positively correlated with the
percentage of income spent online in the last quarter. While demographic factors account for less of the variance
than in the first regression, the percentage is statistically significant. Males spend 0.361% more of their total
quarterly income online than do women, an increase of $40.05 quarterly (actual; Note 7). This finding is
particularly important since it is a far greater spending increase than previously reported for men who purchased
$3.15 more online annually than did women (Lohse et al., 1999). College graduation increases the percentage of
total quarterly income spent online by 0.24% (model derived; Note 8) above a high school degree. These
findings track previous results closely and align with the demographic hypotheses. Targeting an audience that is
male and well educated should produce more revenue and allow for highly efficient marketing.
The influences of the socioeconomic determinants are somewhat different in the second regression. There is a
significant, though very small, negative correlation between household income and percentage of income spent
online because households with lower incomes tend to spend a larger percentage of income online, even though
the actual amount spent online is less than that spent by higher income households. Possession of an online
deferred payment account is a strong positive determinant of the percentage of income spent online in the last
three months. In contrast, possession of an online payment account such as PayPal™ is not a significant
determinant. Online purchasers using deferred payment accounts spend 1.79% more income per quarter.
Although economic data are not available, online deferred purchases are relatively large (usually requiring a
purchase of at least $99) and possibly account for an increase in a user’s online spending. To generate more
revenue, these results suggest that online retailers should provide more methods for buyers to establish and
utilize online deferred payment accounts, focusing less on PayPalTM and other similar payment systems.
Internet usage determinants encompass several correlated variables with the percentage of income spent online,
positive determinants are high Internet usage and communication via Instant Messenger while email is a negative

Published by Canadian Center of Science and Education 63

www.ccsenet.org/ibr International Business Research Vol. 3, No. 4; October 2010

determinant. The only significant information retrieval variable is obtaining stock quotes, which is positively
correlated with the percentage of income spent online in the last three months. Within both models there are
significant correlations with participation in online auctions but not with posting classified ads. The strong
positive correlations observed here suggest that heavier Internet users, stock owners and those who participate in
online auctions are more likely to spend a higher percentage of their income online, providing target markets for
online retailers.
In the second model, all entertainment and social networking variables are determinants of percentage of income
spent online. Podcasting and online dating are positive determinants while online gaming, visiting social network
sites and blogging are negative determinants. These results suggest that online retailers could more profitably
shift their advertising to podcasts, communication providers and online dating services.
While perception of a product as new and innovative is a positive determinant, good value is a significant
negative determinant. Given these results, product lines that are novel or evolve frequently should be highlighted
as such. E-retailers’ marketing should align with consumers’ perceptions and the Internet provides unparalleled
flexibility for retailers to update websites to coincide with their target audience’s perceptions and desires.
Alternative activities yield little explanatory power in either model. The lack of significant correlations between
weekly television hours and percentage of income spent online again suggests electronic retailers should steer
marketing projects away from television and towards more profitable advertising channels such as podcasts or
other online venues that reach buyers more directly. Given that it is not currently possible to selectively bypass
ads in video viewed online, advertisers have the opportunity to target marketing to this captive audience or even
personalize advertising using viewers’ search profiles. Although television viewing online was not a variable
studied here, it should be incorporated into future research because of its rapid expansion (a sevenfold increase
between 2006 and 2007) due to increased broadband adoption coupled with the virtually limitless archived and
contemporaneous media available from television networks or free from third parties (Pew Internet, 2009a,b).
Model 2 Using Partitioned Data. The data set is partitioned to better understand the influence of gender, age and
income, the only variables that show non-normal distributions. (Nationally representative information may still
be obtained by weighting partitioned categories). The first partitioned data set is stratified by gender; then each
data set is applied to the second model (Table 2). The results for men (Model 2B) show much more variance
explained (31.5%) than for women (Model 2C, 8.6%). The models reveal very different determinants of online
spending for men and women, but the reasons for these differences are beyond the scope of this study. The only
commonalities across gender are in the impacts of income (a small, negative effect) and possession of an online
deferred account, though the coefficient for men is almost six-times greater than that of women.
For males, educational attainment, deferred online payment, use of instant messaging, downloading software,
online dating, and perception of new and innovative products are all significant determinants, with a deferred
online payment account providing most of the predictive power (Table 4). Interestingly, significant negative
determinants are use of email, online gaming and blogging. Men appear to be strongly attracted by novel
products and are not value shoppers. Online gaming is a highly significant negative determinant only for men,
aligning with the rest of this data suggesting that men use online shopping as a form of entertainment. Deviating
from the overall data set, online male shoppers show a significant negative correlation between percent of
income spent online and hours watching television.
With its lower explained variance, the regression model for women has fewer significant determinants but
unearths some key gender-related differences. The determinants of women’s online spending behavior support
Bellman et al.’s (1999) theory of time starvation. For women only, work hours and having a high-speed
connection are positive determinants. Similarly, none of the entertainment usage or social networking variables
is significant. In addition to being more convenience-oriented, women who spend more online do not appear to
be drawn to new or innovative products. Sophisticated Internet users, they participate in online auctions and
appear to be responsible for the significance of stock quotes as a significant determinant for Model 2A.
Further partitioning of the data for men by income and age (Models 2D, 2E, 2F) yields improvements in the
percentage of variability explained (Table 2). In Model 2D, for men with household incomes less than $35,000,
66.7% of the variability is explained by this model. Positive determinants are educational attainment, deferred
payment, Instant Messenger, online auctions, classified ads, and new and innovative products (Table 4).
High-speed connection, product research and weekly work hours are negative determinants. A college graduate
at this income level, on average, spends 1.15% (model derived) more income online quarterly than someone with
just a high school degree. The most important factor, the variable with the largest coefficient, is the possession of
an online deferred payment account. This result is not surprising and again points to the attention online retails
should give to such advertising venues. These results illustrate a profitable target market for retailers: men with
household incomes less than $35,000 but with a higher education seem to be novelty-oriented. Men with a

64 ISSN 1913-9004 E-ISSN 1913-9012

www.ccsenet.org/ibr International Business Research Vol. 3, No. 4; October 2010

household income less than $35,000 spend 1.17% (actual) of their income online each quarter, a higher
percentage than those earning more.
Although individuals in the highest income bracket (household income of greater than $100,000) spend more
online than those in the lower income brackets, their percentage of income spent online is lower, 0.75% (actual).
The partitioned data set of males with household incomes greater than $100,000 (Model 2E) explains 29.1% of
the variance in online spending (Table 2). Similar to lower earners, educational attainment, online auctions and
novel products are positive determinants but not weekly work hours (Table 4). Positive Internet usage
determinants are quite different for this economic group; high speed connection, weekly usage, travel research,
and online dating. Not surprisingly, possession of an online deferred account is not a determinant of online
spending for these highest earning men and good value is actually a negative determinant for this group. Online
gaming is also a negative determinant. Given the very distinct results for high-income men, e-retailers probably
require a specialized approach, using only select Web 2.0 channels to attract the business of these high earners.
This group was the smallest of any examined (n=226), and it is possible that a model using their actual income
figures might provide additional insight into the online consumer behavior of this important marketing target.
Online retailers need to rely on models that explain large amounts of variance in their target market. Model 2F,
applied to men aged 45-54, accounts for 79.1% of the variance at a highly significant level despite the lack of a
significant correlation with household income. In this age group, positive determinants are downloading software,
instant messaging, online dating, and most significantly, having an online deferred account and products
perceived as new and innovative (Table 4). The highest observed coefficients in this model are again associated
with online deferred payment. Negative determinants are high-speed connection, product research, social
networking sites, online gaming and weekly television hours.
While the R2 values for Models 1 and 2 are relatively low (23.7% and 19.4%, respectively), both regression
equations were statistically significant (p value < 0.001) and our determinant analysis relies only on variables
that show statistical significance based on at least a p value ≤ 0.05. Admittedly, the extent to which unknown,
confounding variables affect the model’s results are unknown and beyond the scope of this study. Additionally,
several key variables, both demographic and socioeconomic, are missing from the data set and limit its
interpretation. More complete information about survey participants would allow for a richer and more thorough
analysis.
Nevertheless, the large number of survey participants allows the data to be partitioned in order to further explore
purchasing behaviors. An improvement in the explained variance of Model 2 (spending) was found for men but
not for women, suggesting that the inclusion of other, unstudied variables in a regression analysis may provide
more explanatory power of female online shopping behavior. Partitioning the data for men enhanced the
explanatory power of Model 2 for spending by men with household incomes less than $35K and of ages 45-54.
Knowledge of the causality and process between the demonstrated determinants and online spending would be
necessary to translate these linkages into more effective marketing strategies.
5. Conclusions
According to the Internet Advertising Revenue Report (2009), Internet advertising grew 10.6% in 2008. In the
midst of an economic downturn, and in a year in which cable television advertising was the only other category
of advertising to grow, this clearly indicates the importance and potential of the online venue (Interactive
Advertising Bureau 2009). Recognizing its importance, the issue then becomes how e-retailers can most
effectively utilize Internet advertising, especially in the vastly growing and virtually unstudied Web 2.0 realm.
The future growth and success of e-commerce may rest on the utilization of Web 2.0 platforms to advertise to
specific niche users. These new conduits to potential buyers allow for the most efficient marketing to date by
incorporating highly personalized and rapidly dynamic advertising based on personal interests and consumer
history. This study establishes that online auctions, travel research and podcasts are strong determinants of online
purchasing; they provide easily accessible advertising channels. In particular, the podcast audience is a largely
untapped, growing market that may prove lucrative for online firms. Podcast consumers are at least 50% more
likely than non-consumers to have made an online purchase in the past week, are avid consumers of other
communication technology and are active social networkers (Webster 2008). Because possession of an online
deferred payment account was the strongest determinant of spending for all groups except the highest earning
men, opening a deferred payment account should be encouraged by advertising on highly correlated forums for
new and innovative products, online auctions, downloadable software and podcasts.
Drawing on the results of this analysis, electronic retailers may target their efforts and consider the cost-benefit
analysis of each potential advertising project. Increasing online spending in the higher income brackets,
particularly by men, is crucial to the growth of electronic commerce. This demographic is well versed in

Published by Canadian Center of Science and Education 65

www.ccsenet.org/ibr International Business Research Vol. 3, No. 4; October 2010

technology and has a higher than average educational attainment. It is speculated that increasing online spending
in this group can be achieved through selective online advertising integrated with preferred online activities such
as online auctions and possibly podcasting. Their technological sophistication allows them to block many
traditional advertising methods, so advertising via podcasts and online television programs could effectively
target this attractive demographic (Nesbitt 2008).
The models presented here can also provide actual revenue estimates. For example, a college graduate on
average would spend 0.24% more income online than does someone with just a high school degree and
increasing time online of the highest male earners by 20 hours boosts spending by an average of 0.20%, or
$55.00 per quarter for a user that makes $100,000 per year. If online retailers could increase the percentage of
income spent online by the highest income bracket, their revenues would increase significantly. Projected
revenue increases can then be weighed against planned advertising outlays; the resulting cost-benefit analyses
are expected to reduce marketing costs. The possibilities for targeting niche users and attraction of males to
novel and innovative products have implications for website design as well.
Nearly a decade has passed since a representative population of online buyers was last studied in detail, and this
research provides a more focused picture of the current American online buyer. Women are achieving parity as
online shoppers as their technological savvy increases; in some cases, women utilize the Internet at a slightly
higher rate than men for professional information or enrichment (Pew Internet, 2008b). However, analysis of
current data provides evidence that women still appear to be time-starved purchasers. There is a small but
significant increase in the amount purchased online by women as their work hours increase. Women with
high-speed connections, who research products and who participate in Internet auctions spend more online. By
far, though, the largest effect on spending (for both women and men) comes from having an online deferred
payment account.
Although the Internet can provide novel channels to reach perfectly targeted niche users, the ability to integrate
Web 2.0 activities into online marketing remains largely unexplored. The potential to customize online content
and marketing for specific demographics, including by gender, by Internet needs and by online interests, is
particularly important given the distinct spending patterns for men and women and the markedly different
behavior of men in the highest income bracket. Marketing to new cultural or underserved ethnic groups is
another route that is expected to greatly expand future e-commerce, but specific websites and advertising must
be developed to accommodate cultural differences in online perceptions and types of Internet usage (Chau et al.,
2002; Singh et al., 2008). With a better understanding of contemporary American online behavior, this study
points to directions for future research in order to utilizing new and robust marketing platforms to reach the
growing base of people who utilize the Internet for entertainment.
Acknowledgment
We thank Lauren Hannah for her assistance with the categorical statistics and Dr. Jo Ellen Hose for her editorial
suggestions. We are also grateful to the Department of Economics and Business at Colorado College for
financial assistance in purchasing the data.
References
Bakos, J.Y. (1997). Reducing buyer search costs: implications for electronic marketplaces. Management Science,
43(12), 1676-1692.
Bakos, Jannis. (2001). The emerging landscape for retail e-commerce. Journal of Economic Perspectives,15(1),
69-80.
Bellman, S., G.L. Lohse & E.J. Johnson. (1999). Predictors of online buying behavior. Communications of the
ACM, 42(12), 32-38.
Bhatnagar, A., S. Misra & H.R. Rao. (2000). On risk, convenience, and Internet shopping behavior.
Communications of the ACM, 43(11), 98-105.
Brown, J.R. & A. Goolsbee. (2002) Does the Internet make markets more competitive? Evidence from the life
insurance industry. Journal of Political Economy, 110(3), 481-507.
Brynjolfsson, E., Y. Hu & M.D. Smith. (2003). Consumer surplus in the digital economy: estimating the value of
increased product variety at online bookstores. Management Science, 49(11), 1580-1595, 2003.
Chau, P. Y. K., M. Cole, A. P. Massey, M. Montoya-Weiss & R.M. O'Keefe. (2002). Cultural differences in the
online behavior of consumers. Communications of the ACM, Vol. 45(10), 138-143.

66 ISSN 1913-9004 E-ISSN 1913-9012

www.ccsenet.org/ibr International Business Research Vol. 3, No. 4; October 2010

Comiskey, D. (2006). Report: 2006 online retails sales to hit $100 billion. [Online] Available:
http://www.e-commerce-guide.com.
Cummings, J.N. & R Kraut. (2002). Domesticating computers and the Internet. Information Society, 18(3),
221-231.
Donthu, N. & A. Garcia. (1999). The Internet shopper. Journal of Advertising Research, 39(3), 52-58.
Interactive Advertising Bureau. (2009). IAB Internet advertising revenue report. [Online] Available:
http://www.iab.net/media/file/IAB_PwC_2008_full_year.pdf.
Gelman, A & Hill, J. (2006). Data Analysis Using Regression and Multilevel/Hierarchical Models. Cambridge,
UK: Cambridge University Press.
Jupiter Research. (2007). [Ipsos online insight survey (06/07)]. Unpublished raw data.
Korgaonkar, P.K. & L.D. Wolin. (1999). A multivariate analysis of web usage. Journal of Advertising Research,
39(2), 53-68.
Lee, H.G, S.C. Lee, H.Y. Kim & R.H. Lee. (2003). Is the Internet making retail transactions more efficient?:
comparisons of online and offline CD retail markets. Electronic Commerce Research & Applications, 2(3),
266-277.
Li, H., C. Kuo & M.G. Russell. (2006). The Impact of Perceived Channel Utilities, Shopping Orientations, and
Demographics on the Consumer’s Online Buying Behavior. Journal of Computer-Mediated Communication,
5(2).
Lohse, G.L., S. Bellman & E.J. Johnson. (1999). Consumer buying behavior on the Internet: findings from panel
data. Journal of Interactive Marketing, 14(1), 15-29.
Menard, S. (2002). Applied Logistic Regression Analysis (2nd ed.). Thousand Oaks, CA: Sage Publications.
Nie, N.H. & D.S. Hillygus.(2002). Where does Internet time come from? A reconnaissance,” IT & Society, 1(2),
1-20.
Nielsen. (2008). Nielsen’s three screen report. [Online] Available:
http://www.nielsen.com/pdf/3_Screen_Report_May08_FINAL.pdf.
Nielsen//NetRatings. (2006). Over three-fourths of U.S. active Internet users connect via broadband at home in
November, according to Nielsen//NetRatings. [Online] Available: http://www.nielsen-netratings.com.
Nesbitt, A. (2008). The podcast consumer in 2008. [Online] Available:
http://www.digitalpodcast.com/podcastnews/2008/05/30/the-podcast-consumer-in-2008/.
Pew Internet, Inc. (2008a) Demographics of Internet users. [Online] Available:
http://www.pewinternet.org/trends/User_Demo_2.15.08.htm.
Pew Internet, Inc. (2008b) “Latest trends: usage over time. [Online] Available:
http://www.pewinternet.org/trends.asp.
Pew Internet, Inc. (2009a). Internet activities. [Online] Available:
http://www.pewinternet.org/Data-Tools/Download-Data/~/media/infographics/Trend%20Data/January%202009
%20updates/Internet%20Activities%20-%20all%20-%201%206%2009.jpg.
Pew Internet, Inc. (2009b). The mobile difference. [Online] Available:
http://www.pewinternet.org/Reports/2009/5-The-Mobile-Difference--Typology/2-Introduction/3-The-Typology-
and-Longitudinal-Analysis.aspx?r=1.
Singh, N., D.W. Baack, S.K. Kundu & C. Hurtado. (2008). U.S. Hispanic consumer e-commerce preferences:
expectations and attitudes toward web content. Journal of Electronic Commerce Research, 9(2), 162-175.
Susskind, A.M. (2006). Electronic commerce and world wide web apprehensiveness: an examination of
consumers’ perceptions of the world wide web. Journal of Computer-Mediated Communication, 9(3).
Swaminathan, V., E. Lepkowska-White & B.P. Rao. (1999). Browsers or buyers in cyberspace? An investigation
of factors influencing electronic exchange. Journal of Computer-Mediated Communication, 5(2).
Swinyard, W.R. & S.M. Smith. (2003). Why people (don’t) shop online: a lifestyle study of the Internet
consumer. Psychology & Marketing, 20(7), 567-597.
U.S. Census Bureau. (2010). Estimated quarterly U.S. retail sales (not adjusted): total and e-commerce. [Online]
Available: http://www.census.gov/mrts/www/data/html/10Q1table4.html.

Published by Canadian Center of Science and Education 67

www.ccsenet.org/ibr International Business Research Vol. 3, No. 4; October 2010

U.S. Census Bureau. (2009). E-stats - measuring the electronic economy. [Online]
Available:http://www.census.gov/eos/www/ebusiness614.htm
WebSiteOptimization.com. (2008). July 2008 bandwidth report. [Online] Available:
http://www.websiteoptimization.com/bw/0807/.
Webster, T. (2008). The podcast customer revealed: The Arbitron/Edison Internet and multimedia study,” Edison
Media Research.
Zhou, L., L. Dai & D. Zhang. (2007). Online shopping acceptance model – A critical survey of consumer factors
in online shopping. Journal of Electronic Commerce Research, 8(1), 41-62.
Notes
Note 1. Heterosckedasticity was not present in the results; the weighting variable is included to make the mean
data and standard error for each user representative of data for the online population.
Note 2. The weight variable is based on 12 demographic variables. The idea behind the weight is to ensure that
the sample is representative of the US online adult population, according to Jupiter's definition of what the
demographics of that population are. Respondents are invited to take the survey based on specific demographic
quotas, and then the weight is applied to the data to ensure that the distributions are precisely in line with the
same demos.” (Jupiter Research 2007)
Note 3. In this survey effort, Jupiter Research worked with its research partner, Ipsos Insight…Ipsos Insight is
one of the largest market research companies in the US and maintains a general research panel of 400,000
households. Ipsos Insight also has access to the Ipsos US online panel, which comprises two million Internet
users.” (Jupiter Research 2007)
Note 4. Predicted values are determined by applying the participant’s survey answers (such as age, income,
high-speed connection, etc.) to the coefficients from the first regression.
Note 5. This describes an online user who is 40 years old, a college graduate, makes $35,000 per year, works 40
hours per week, uses the Internet 5 hours per week, watches television 10 hours per week, owns a credit card,
has a high-speed Internet connection and of the activities surveyed, uses only email and search engine.
Note 6. This describes an online user who is 40 years old, a college graduate, makes $100,000 per year, works 40
hours per week, uses the Internet 20 hours per week, watches television 20 hours per week and uses the Internet
for every variable surveyed.
Note 7. Actual: these values are based on descriptive statistics, in this case they are the actual differences
between men and women’s online spending, they are derived directly from the dataset, not from the model’s
results.
Note 8. Model derived: these values are obtained through the model by altering the specified variable (education)
and holding all others constant.

68 ISSN 1913-9004 E-ISSN 1913-9012

www.ccsenet.org/ibr International Business Research Vol. 3, No. 4; October 2010

Table 1. Description of Variables in Models and Key Statistics

DEPENDENT VARIABLE DEFINITION OF DEPENDENT VARIABLE
Has purchased a good or service online within the last year (1) or not (0)
Purchase
y1 n = 3580; 72% yes, 28% no
Percentage of income spent via online purchasing in the last three months (the amount spent online in the last
three months [median of bracketed range] by household income for the last three months [median of bracketed
Spending
range divided by four].
y2
n = 2539; mean [SD] = 0.682% [1.603%]
CATEGORY AND NAME OF
DEFINITION OF INDEPENDENT VARIABLE
INDEPENDENT VARIABLE
Demographic
Median age of each bracket is applied to data set: 21 (18-24), 29.5 (25-34), 39.5 (35-44), 49.5 (45-54), 55 (55+).
x1 Age Mean [SD] = 37.1 [13.9]
Male (0) or female (1).
x2 Gender 50% male, 50% female
Highest level of education the respondent completed
(1) Grade school (2) Some high school (3) Graduated high school (4) Some College (5) College graduate (6)
Educational
Post-graduate degree
Attainment
The average respondent has completed some college. 6% some high school, 26% graduated high school, 28%
x3
some college, 28% graduated college, 12% post-graduate degree
Socioeconomic
Median value from the bracketed range applied to the data set
Household Income $ 35K (<$35K), 39.5K (35K-44K), 52K (45K-59K), 67K (60K-74K), 87.5K (75K-100K), 100K (100K+).
x4 Mean [SD]= $59,130 [$29,891]
High-speed Dial-up or no connection at home (0) or high-speed (1).
x5 Connection 63% high-speed connections, 34% dial-up connections, 2% without connection and 1% did not know their
connection type
Possession of credit card or debit card that can be used in online purchasing (1) or does not own any form of
Credit or Debit Card online payment method (0).
Ownership 84% own payment card, 16% without payment card. (This variable is not used in Model 2 since an insignificant
x6
number of participants did not possess a payment card )
Has online payment account such as PayPal™ (1) or not (0).
x7 Online Payment 37% yes, 63% no
Online Deferred Has online deferred payment account (1) or not (0).
x8 Payment 5% yes, 95% no. (This variable is not used in Model 1 because it assumes user has purchased online)
Internet Usage
Average hours spent online weekly by the user.
Weekly Internet Usage
x9 Mean [SD] = 19.3 [21.7]
Communication
User has sent or received email in last month (1) or not (0).
x10 Email 90% yes, 10% no
User has used AOL Instant Messenger, ICQ, Yahoo or MSN Messenger, or similar instant messaging services in
the last month (1) or not (0).
x11 Instant Messenger
37% yes, 63% no
Information Retrieval
User has researched product or services online in the last year (1) or has not (0).
x12 Product Research 52% yes, 48% no
User searched for information using a search engine within the last month (1) or has not (0).
x13 Search Engine 78% yes, 22% no
User has downloaded software programs for their personal computer in the last year (1) or not (0).
Downloaded Software
x14 25% yes, 75% no
User has investigated travel arrangements (availability or pricing) online in the last month (1) or not (0).
Investigated Travel
x15 41% yes, 59% no
User has checked stock quotes online in the last month (1) or not (0).
Stock Quotes
x16 14% yes, 86% no
Shopping Behavior
User has sold or bid for products in an online auction within the last month (1) or has not (0).
x17 Online Auction 21% yes, 79% no
User posted classified ads online (e.g. Craig’s List, AutoTraderTM, etc.) in the last month (1) or has not (0).

Published by Canadian Center of Science and Education 69

www.ccsenet.org/ibr International Business Research Vol. 3, No. 4; October 2010

x18 Classified Ads 8% yes, 92% no

Entertainment
User has listened to or downloaded a podcast within the last month (1) or not (0).
x19 Podcast 12% yes, 88% no
User played games online such as action games, fantasy, flight simulators, etc. in the last month (1) or has not
(0).
x20 Online Gaming
27% yes, 73% no
Social Networking
User has visited social networking sites like MySpaceTM, FacebookTM, etc. in the last month (1) or has not (0).
x21 Social Networks 36% yes, 64% no
User has used an online dating service or viewed personal ads in the last month (1) or has not (0).
x22 Online Dating 15% yes, 85% no
User has read a blog in the last month (1) or has not (0).
x23 Blogs 26% yes, 74% no
Product Perceptions
User prefers to buy products that are a good value for the money (1) or not (0).
x24 Good Value 78% yes, 22% no
User prefers to buy products that are new and innovative (1) or not (0).
x25 New/Innovative 16% yes, 84% no
Alternative Activities
Average hours user works weekly.
x26 Weekly Work Mean [SD] = 29.6 [27.2]
Average hours user watches television weekly.
x27 Weekly Television Mean [SD] = 16.9 [17.6]

Table 2. Results from Regressions

Adjusted F Statistic Number of Degrees of
Model R-Squared (p value) Observations Freedom
1115.76
Model 1 Purchase 0.237 (p<0.001) 3580 3553
23.277
Model 2 Spending 0.194 (p<0.001) 2539 2512
23.239
Spending by Men 0.315 (p<0.001) 1213 1187
5.473
Spending by Women 0.086 (p<0.001) 1197 1171
Spending by Men with 27.000
Household Incomes < $35K 0.667 (p<0.001) 312 287
Model 2 with Spending by Men with 7.028
Partitioned Data Household Incomes > $100K 0.291 (p<0.001) 226 201
45.250
Spending by Men Aged 45-54 0.791 (p<0.001) 281 256

Table 3. Results from Model #1 & #2 (Demographic Variables Only).

Adjusted F Statistic
Number of Observations Degrees of Freedom
Model R-Squared (p value)
123.242
Model 1 Purchase 0.033 (p<0.001) 3580 3576
17.415
Model 2 Spending 0.021 (p<0.001) 2539 2535

Model 1 Model 2
Purchase Spending
Constant -1.419*** 3.341***
Demographic
Age 0.012*** 0.001
Gender 0.217** -1.272**
Education 0.316*** 0.235*

70 ISSN 1913-9004 E-ISSN 1913-9012

 www.ccsenet.org/ibr International Business Research Vol. 3, No. 4; October 2010

* 0.01 < p value ≤ 0.05

** 0.001 < p value ≤ 0.01
*** p value ≤ 0.001

Table 4. Coefficients and Significance from Regressions Using Partitioned Data

Model 1 Model 2
2A 2B 2C 2D 2E 2F
Spending Spending Spending
Purchase Spending Spending Spending by by Men by Men by Men
by Men Women <$35K > $100K 45-54

Constant -3.683*** 0.684*** 0.255 0.750*** -1.033 -1.725** 0.481

Demographic
Age 0.006 0.002 0.007 -0.003 -0.008 0.009 
Gender 0.290** -0.114*     
Education 0.069 0.120*** 0.221*** -0.015 0.573*** 0.259** 0.079
Socioeconomic
Income 0.128*** -8E-006*** -8E-006*** -6E-006***   -4E-006
High-Speed Connection 0.183 0.096 0.153 0.205*** -0.429* 0.553** -0.330*
Credit Card 1.279***      
Online Payment 0.637*** -0.048 -0.155 0.040 -0.065 -0.179 0.029
Online Deferred Payment  1.797*** 2.973*** 0.498*** 4.627*** -0.028 3.614***
Internet Usage
Weekly Usage 0.003 0.005** 0.001 0.003 0.003 0.011* 0.001
Communication
Email 0.588*** -0.247* -0.516** 0.158 -0.237 0.211 -0.036
Instant Messenger -0.133 0.235*** 0.433*** -0.085 0.703*** -0.212 0.478*
Information Retrieval
Product Research 0.471*** -0.090 -0.101 0.105 -0.542*** -0.147 -0.524**
Search Engine 0.588*** 0.125 0.155 0.046 -0.391 0.093 0.119
Download 0.273* 0.112 0.211* -0.098 0.125 0.050 0.645***
Investigate Travel 0.426*** -0.107 -0.148 0.057 -0.062 0.369* 0.054
Stock Quotes 0.215 0.162* 0.066 0.473*** -0.339 0.230 0.081
Shopping Behavior
Online Auction 0.877*** 0.125* 0.125 0.197** 0.513** 0.696*** -0.134
Classified Ads -0.008 -0.084 0.029 -0.226* 0.798** -0.205 0.230
Entertainment
Podcast 0.403* 0.149* 0.086 0.089 -0.106 0.254 -0.131
Online Gaming -0.090 -0.233*** -0.343*** -0.089 -0.192 -0.427* -0.513*
Social Networking
Social Networks 0.039 -0.126* -0.129 -0.149 -0.345 -0.296 -0.743***
Online Dating 0.312 0.310*** 0.336*** 0.057 0.263 0.483* 0.689***
Blogs 0.318 -0.020*** -0.253** 0.100 0.260 0.101 -0.329
Product Perception
Good Value 0.111 -0.281*** -0.320** -0.298*** 0.184 -0.656*** 0.304
New/Innovative 0.391** 0.505*** 0.667*** 0.153 0.753*** 0.439*** 1.314***
Alternative Activities
Weekly Work -0.001 0.001 -0.001 0.003** -0.011* -0.006 -0.003
Weekly Television 0.005 -0.003 -0.007* 0.000 -0.006 0.002 -0.010
 The variable Online Deferred Payment was not included in Model 1 because it is assumes that the user has made an online
purchase. The variable Credit Card was only used in Model 1 because there were an insignificant number of users in the dataset
who lacked a credit card .
 These variables were not included when the data were partitioned by gender and/or income.
* 0.01 < p value ≤ 0.05,
** 0.001 < p value ≤ 0.01,
*** p value ≤ 0.001

Published by Canadian Center of Science and Education 71