There has been phenomenal Internet penetration in India in recent years, from just 4% in 2007 to nearly 50% in 2020 [1
]. This has led to a substantial increase in e-retail sale. The e-retail market itself is estimated to grow from US$
24 billion in FY 2019 to US$
98 billion in FY 2024 [2
]. The sector is expected to continue growing and attracting attention of e-retailers across the globe. However, at present, 80% of the population who shop online are young people (below 35 years) while other consumer segments’ acceptance of e-shopping as a purchasing channel is low. Thus, the percentage of online sales in terms of total retail sales was only 1.6% in India, versus over 15% for China and around 14% globally and hence e-retail is considered to be still in a nascent stage in India [3
]. In order to increase the online sales, it is important to make middle aged and older population also to shop online. Ignoring this fact, most extant e-shopping research studies in India have concentrated on understanding the motivation of young people for e-shopping (meta-analysis by [4
]), rather than identifying the barriers that prevent others from shopping online [6
], which is the purview of current study.
Perceived risk is considered the most important psychological state that negatively influences one’s e-shopping behavior as shown by many studies [9
], meta-analysis by [19
]. However, most of the earlier studies have been conducted in developed countries [15
]. The risk perception towards any new technology is found to be high in developing countries, especially in countries like India with a collectivistic culture [9
]. People from countries like India are risk-averse and hence are slow in the adoption of any technology. Hence, e-shopping researchers have noted that to make these people shop online, it is important for retailers to focus on reducing risks rather than concentrating on benefits when setting up their online presence [12
]. E-shopping researchers have emphasized that there exists a knowledge gap on the factors that act as barriers of adoption of e-shopping technology in developing countries [5
]. Moreover, while several studies have highlighted the importance of considering the multi-dimensionality of perceived risk [15
], few studies have been conducted taking the various dimensions into account [6
]. Studies that have assumed perceived risk to be a uni-dimensional construct have observed insignificant effect of perceived risk on behavioral intention (BI) in collectivistic countries; they have reported that it is important to further disintegrate the risk construct to gain better understanding [16
]. Hence this study considers risk as a multi-dimensional (ten) construct (Table 1
) and studies the impact of various risk perceptions on e-shopping behavioral intention (BI) in the Indian context.
Researchers have noted that to better understand the effect of perceived risk on BI, it is crucial to consider the moderating roles of various socio-demographic factors (meta-analysis by [30
]). Perception towards risk is dependent on an individual’s socio-demography, culture, and traits. People with different socio-demographic and cultural backgrounds encounter different situations in life [8
]. Different types of risks are formed with respect to a particular situation encountered by an individual [31
]. Moreover, time and psychological pressures encountered at different situations have been found to have a significant effect on risk perceptions, and the corresponding risk aversion levels also differ [34
]. Past researchers have investigated the effect of various socio-demographic factors on perceived risk. However, studies that focus on understanding the differences in e-shopping behavior based on family life cycle (FLC) stages are limited [8
]. There is a strong evidence from past literature that most situational and behavioral changes in life happen due to the FLC stages of an individual rather than his/her age. Changes in FLC stages result in differences in an individual’s financial, time and psychological pressures and involvement with purchases [39
]. Against this backdrop, we expect that there would be differences in the perceptions of risk dimensions and their interplay with BI towards e-shopping sites, as an individual transcends the various FLC stages.
This study addresses two potential segments that are often ignored in e-shopping studies. First, most e-shopping studies on perceived risk are conducted among young people, while the pattern of risk perception among the middle-aged or older adults often considered lucrative market segments with higher disposable income—has been least explored [46
]. For e-shopping to expand and prosper, it is important to concentrate on people of all ages and at different stages of life [5
]. Given that e-shopping is an integral part of today’s economy, understanding the e-shopping behavior across all age groups becomes important, as the adoption of technology by older adults plays a vital role in economic development [49
]. Hence to understand the effect of different demographic variables on online behavior, a further split age-based research on technology adoption is essential [26
]. This study acknowledges that there are barriers that hinder people of different ages from shopping online and explores this further.
Another potential segment is women e-shoppers. Fewer women are online in India than in other BRIC countries. The dearth of studies on e-shopping behavior among Indian women has also been an outcome of the nascency of e-commerce in India [51
]. Women perceive higher risks in e-shopping than men [16
] especially in developing countries like India [21
]. Given that in most cultures, women assume the responsibility of household shopping and purchase decisions [53
] and FLC stages have a greater effect on women than men [56
], there is a need to explore the role of perceived risk in deterring Indian women of different FLC stages from participating in e-shopping activities. In addition, women are identified to spread word of mouth regarding their shopping experience than men. Hence the ability of women to attract more customers towards online shopping sites is higher than men. In addition, at the time of purchase, men are more direct and only buy at the time they need something. On the other hand, women tend to try more items than men and are more likely to make impulse purchases. Hence by targeting women it is possible to increase online sales, in-turn profit. Moreover, young women are more inclined towards e-shopping and there are many studies that have analyzed the e-shopping behavior of this section of the young population. Our study focuses on married women, i.e., women who are a part of a marital family, with and without children, and analyses the changes in their perception of e-shopping risk, which in turn, affects their BI to shop online. Since risk perception toward new technology negatively influences e-shopping behavior, and the problem is particularly bad in developing countries like India with collectivistic cultures, it pays off to try to better understand how risk perceptions influence e-shopping behavior to mitigate the problem. There are large individual differences in perceived risk, and to better understand individual factors behind risk perceptions, the current study employs the FLC.
2. Objectives of the Study
This study, as a pioneering effort, attempts to fill the knowledge gap in the area of e-commerce adoption, identified through a search of literature, by understanding the effect of different dimensions of perceived risk on e-shopping purchase intention among Indian women of different FLC stages. To operationalize perceived risk, the study uses the theory of perceived risk (TPR) proposed by [57
]. The theory states that a customer, while buying any product would perceive some amount of risk. They believe that their purchasing actions would lead to consequences that cannot be anticipated with anything approximating certainty, and some of the consequences are likely to be unpleasant. Hence, the consumer’s choices are divided into risk-increasing or risk-decreasing behavior. Bauer initially proposed perceived risk but did not include specific types. Later researchers have considered perceived risk as a combination of several dimensions [58
]. Under the new Internet environment, as scholars adjusted the model according to the development of technology, economy and society, more dimensions were identified. However, there is no agreement on the dimensions of perceived risk.
Researches on perceived risk have divided it into different dimensions according to different situations. Not all dimensions of perceived risk have been found to have significant effects on consumer’s behavior [59
]. Thus, there is a need to examine the effect of different dimensions of perceived risk on a consumer’s behavior in sharing economy [27
In this study, we have considered ten dimensions of perceived risk (Table 1
)–financial risk, performance risk, time-loss risk, privacy risk, delivery risk, social risk, after-sales service risk, source risk, psychological risk, and physical risk [60
], and have identified the risk dimensions affecting e-shopping behavioral intention among women of different FLC stages. To operationalize the FLC stages, a customized Indian FLC stages model comprising nine FLC stages, proposed by [36
] has been used (Table 2
Furthermore, the researchers that applied TPR to understand e-shopping risk perception highlighted that gender plays a crucial role in affecting e-shopping risk perception; women are identified to be risk-averse than men [61
]. Hence, this study attempts to test how the perceived risk perception influence Indian women’s e-shopping behavior as they transcend through different FLC stages.
With this background, we arrive at the following research questions (Figure 1
Among the ten dimensions of perceived risk (financial risk, performance risk, time-loss risk, privacy risk, social risk, psychological risk, delivery risk, after-sale risk, source risk, physical risk), which dimensions of perceived risks are negatively associated with the purchase intention to shop online, among Indian women?
Does the effect of these risk dimensions on BI differ as women transcend from one FLC stage to the next?
5. Research Methodology
The sample data consisted of female respondents from selected metropolitan cities of India, owing to the high internet diffusion rate in these cities [3
]. To participate in the survey, it was ensured that the respondents met the following criteria: The women respondents were those:
Who have never shopped online, as that would exactly replicate the risks hinder them from shopping online;
Residing in selected metropolitan cities of India (Chennai, Bangalore, Hyderabad, Mumbai, and Pune) and;
Living in nuclear families since the nuclear family system predominantly exists in metropolitan cities of India [133
]. Furthermore, it is believed that time and work pressures are higher in nuclear families than in extended/joint families, which would affect e-shopping behavior.
Who have remained in that particular FLC stage for the past 6 months to ensure buying behavior of that particular stage.
5.2. Data Collection Instrument
This study is non-experimental and quantitative in nature. The study used offline mode for data collection. Standardized self-report questionnaires were used to collect data, which had two sections. It began with the administration of screening question to respondents to ascertain if they have had any prior e-shopping experience. If the answer was no, the respondent proceeded with the questionnaire. Section 1
focused on the demographic profile of the respondent and collected details about family-related information including respondent’s age, spouse’s age and employment status, age of children, information on children living with them or not, to understand the FLC stage to which they belonged. Section 2
captured the perception towards e-shopping perceived risk dimensions and BI. The [134
] scale was used to measure performance risk (three items), financial risk (three items), time-loss risk (three items), psychological risk (three items), source risk (three items), physical risk (three items) and [60
] scale was used to measure social risk (three items), privacy risk (three items), delivery risk (three items), after-sale risk (three items) which consisted of a total of 33 scale items measured on a seven point Likert scale anchored by “strongly disagree” to “strongly agree” [135
]. As the questionnaire (Appendix A
) has not been tested in India, a pilot study was conducted by collecting data from 10 samples across all segments. Following this, full-fledged data collection was made.
5.3. Data Collection Procedure
Data is gathered via offline mode. We met respondents directly at home. We preferred society/apartments as it is easy to meet many respondents at a time. The society/apartment head is contacted and explained with the data collection process and with their support, the hard copies of the samples are distributed to all the houses in the apartments. Gifts and rewards are announced to encourage participation. We requested the samples to participate in the survey staying at home. The questionnaires are collected from their homes after three days.
Data collection was made over a period of four months in five metro-cities, i.e., Chennai, Bangalore, Hyderabad, Mumbai, and Pune, in order to have a wider variety of population. The study sought 100 filled questionnaires from each FLC stage. Respondents were met directly at home/apartments and requested to participate in the survey. The homes and apartments were chosen in such a way that it covered the major areas of the cities. Gifts worth Rs. A total of 30 were announced to all participants. Additionally, a lucky prize worth Rs. 1000 was announced for three participants to improve the participation rate. Respondents were given a time of three days to complete the questionnaire.
Two rounds of data collection were carried out. The first round was based on the convenience sampling method. Homes/apartments located in different parts of the five metro-cities were chosen and around 3500 respondents meeting the study criteria were requested to participate in the survey. Finally, 725 respondents agreed to participate in the survey in Round 1. After excluding incomplete questionnaires and treating responses with a few missing values using the AMOS maximum likelihood method, 673 usable responses were obtained. Of the received responses, 53% of the respondents were in early LC stages (357), and 47% were there in later LC stages (316). To obtain an adequate number of required samples, a second round of data collection was carried out.
The second round was based on both convenience and snowball sampling. A total of 1685 participants were identified through snowball sampling, with the help of the respondents who participated in Round 1, along with convenience sampling. All participants in Round 1 were requested to suggest potential respondents from nearby areas after they completed the survey. With the information received, data was collected from these participants and also through our own search. This resulted in 349 responses, out of which 240 were usable. Among them, 56% belonged to early LC stages (143) and 44% belonged to later LC stages (97).
A total of 913 usable responses were received: 500 were in early LC stage and 413 were in later LC stage. When split based on FLC stages, there were 103 NM, 87 EP, 103 FN(I), 102 FN(PC), 105 FN(SC), 105 FN(TC), 105 FN(AD), 104 FN(AI), 99 EN&SS. This sample size is larger than that used in previous studies [42
]. We then proceeded with data analysis.
5.4. Sample Demographics
shows the sample demographics. The respondents were in the ages of 21–62 years. The average age of respondents was about 43 years. In terms of employment status, 34% of women were employed full time, 25% were part time or work-from-home employees and 41% were housewives. A total of 28% of women had completed their post-graduation, 35% had under-graduation degree, 27% had higher secondary education, and the remaining 10% had below higher secondary education level. This is similar to the observation of previous studies that women who have not shopped online are unemployed and less educated [53
5.5. Data Analysis
As recommended by [137
], a three-step approach was adopted to analyze the collected data.
First, all measurement models were checked for their psychometric properties—the reliability and validity. To do this, Smart PLS 3.0 software package was used, as it is an extensively accepted variance-based, descriptive and prediction-oriented approach in structural equation modelling [138
]. To ensure the data were free from common method bias, Harman’s single factor test was conducted, using SPSS 21.0.
To measure the relationship among the perceived risk dimensions and BI across FLC stages, partial least square structural equation modelling (PLS-SEM) was used as it is an appropriate technique for studies with small sample sizes and models with multiple relationships [139
Multi-group analysis (MGA):
To know if there were significant group differences across FLC stages in the dimensions of perceived risk and BI relationship, PLS-MGA was carried out as it is considered to be the simple but robust method [138
6. Results and Discussion
6.1. Preliminary Tests
To assess the measurement model, the data were checked for reliability and validity. Reliability was checked with composite reliability (CR) values. All the CR values were above 0.7 (Appendix A
), meeting the threshold requirement thus fulfilling the reliability test [140
After the reliability test, the data was tested for its validity—both in terms of convergent and discriminant properties. Convergent validity of the constructs were tested with two values [141
]; (i) factor loadings, and (ii) average variance extracted (AVE). The factor analysis results showed factor loadings of items on the constructs. All items had factor loadings above the recommended level of 0.70, on their corresponding constructs, thus indicating good convergent validity (Appendix A
). Similarly, all the AVEs ranged from 0.676 to 0.807, which exceeded the recommended level of 0.50 (Appendix A—Table A2
). This shows that more than one half of the variances observed in the items were accounted for by their hypothesized constructs. Thus, both conditions for convergent validity were satisfied.
Following this, to check discriminant validity of the data, the shared variances between factors were compared with the AVE of the individual factors [141
]. The result showed that the AVE of the individual factors, were higher than the shared variances between factors, confirming discriminant validity. Further, the factor analysis result showed that factor loading on the respective construct was larger than its loadings on all other constructs, and there was no cross-factor loadings (Appendix A
), indicating good discriminant validity [142
]. Thus, the measurement model demonstrated adequate reliability, convergent validity and discriminant validity.
Then the ‘Harman’s one-factor’ common method bias test was conducted. The un-rotated principal component analysis including eleven factors counts for 37% of the total variance, lower than the cut-off value of 50%, indicating that the data is free from common method bias.
6.2. Structural Model
Once the data satisfied the reliability and validity tests, the structural model was tested to confirm the hypothesized causal relationships among the constructs under study. This study proceeded to test the path significances using a bias-corrected bootstrapping re-sampling technique with 5000 sub-samples [139
]. The path coefficients (beta values) indicated the strengths of the relationships between constructs. Paths with t-values greater than or equal to 1.96 with a significance level of 0.05 [16
] was considered to be significant relationship. R2
value indicates the overall predictability of the model.
Baseline model testing: Overall results
As shown in Figure 2
, the results of PLS analysis showed that, of the ten dimensions of perceived risk, only five dimensions have significant negative effect on BI to shop online. The directions of the relationships (negative effect) are in accordance with the findings observed in other studies [9
], and the meta-analysis by [19
]. Factors in order of relative importance affecting BI to shop online were performance risk, social risk, time-loss risk, psychological risk and source risk while financial risk, after-sale risk, physical risk, delivery risk and privacy risk were found to have insignificant effects [60
]. Checking R2
values, perceived risk predicted 75.7% of intention. Thus, it confirmed the model validity by meeting the reasonable criteria (above 0.19) to confirm the model validity [143
Across all relationships, performance risk was more influential in affecting BI to shop online, than any other risk dimensions (twice as influential as the social risk—the second ranking risk dimension). This is similar to the results observed in several previous studies [24
]. Following this, social risk and time-loss risk were found to play dominant roles. This is also in line with the results observed in past studies [24
]. However, some studies have also shown that social risk and time-loss risk had insignificant effect on BI to shop online in certain developing countries [60
]. However, as far as e-shopping in India is concerned, social risk and time-loss risk continue to play an important role in affecting e-shopping behavior among women. The reason for social risk to show a significant effect may be that women in India buy in groups and enjoy group memberships. When one shops online, one may feel insecure that she may be separated from the group. In a similar vein, the significant effect of time-loss risk on e-shopping BI may be due to the time constraints that Indian women have; due to excessive time spent on taking care of children and household responsibilities than men. Hence losing time may be of great concern to them. Psychological risk was found to play a dominant effect. This may be attributed to the fact that e-shopping is in nascent stage in India and most women are used to traditional shopping formats.
Financial risk was also found to be an important barrier for e-shopping [12
]. The results of this study showed insignificant effect. A similar result was also observed in a previous study conducted in China [144
]. The reason for financial risk being insignificant maybe because customers prefer the cash on delivery mode of payment to online payments in India [2
] due to poor internet security. Studies have noted that the perception towards financial risks differed widely with culture and e-shopping adoption rate [60
]. Privacy risk was found in earlier studies to be a major deterrent affecting e-shopping BI particularly in collectivist cultures [145
]. However, our study showed an insignificant effect of privacy risk on e-shopping BI. Studies have shown that except for expensive products, privacy risk did not have a significant effect on others [29
]. As purchase of low-ticket items is more popular in India than the purchase of expensive goods, privacy risk would not have shown a positive effect. Moreover, earlier studies have shown that women are concerned about privacy and hence take all necessary measures to protect their privacy [130
]. Hence, women would be proactive in protecting their privacy while other risks which are beyond their control would show a stronger effect. Alternately, just mitigating privacy and security risk on e-shopping, would not make one shop online. Privacy and security maybe the basic things expected by e-shoppers. It may be a hygienic factor as noted by Herzberg’s theory, the absence of which would affect adoption of e-shopping, while its presence may not contribute to buying behavior. Physical and delivery risks were found to have insignificant effect in our study similar to the result observed in previous studies [25
6.3. FLC Stage-Wise Model Testing
Across early LC stages, of the ten dimensions of risk, only three dimensions were found to affect people of early LC stages—performance, time-loss and after-sale (Table 4
; Figure 3
For NM women, no other risk, except performance was found to affect e-shopping purchase intention. EP families avoid shopping online due to their strong perception of performance and after-sale service risk. Once children enter into the family, it could be observed that time-loss risk starts to play a much more crucial role than after-sale service risk (Table 4
—higher t-values) in affecting BI to shop online. In [36
], the authors noted that time saving acted as an important reason for these young families with children to shop online. With this result, it is noted that the same factor also acts as a great hindrance to adopt e-shopping. Hence, it could be inferred that time-loss risk may act as a barrier before adoption, but once adopted, e-shopping offers time-saving benefit that the same acts as a strong reason to shop online. However, other risk dimensions like financial, privacy, social, source, psychological, physical, delivery were found to be insignificant in affecting early LC stages’ BI to shop online.
As women reach later LC stages, of the 10 risk dimensions, 5 dimensions—performance, social, source, psychological and after-sale were found to affect e-shopping BI. Performance risk continued to be a strong barrier to BI, among all later LC stages. However, the time-loss risk has been found to lose its effect while psychological risk, source risk and social risk starts gaining importance [25
]. As expected, since these older families are not used to e-shopping, psychological risk of adopting a new channel and if at all adopting, confusions about the reliable site (i.e.,) source risk, all act as hindrances among these later LC stages. However, this was not applicable to all later LC stages, only 2 groups—FN(AD) and EN&SS were found to be affected by source risk, while it was insignificant among groups FN(TC) and FN(AI). EN&SS stages perceived after-sale service risk as these people are new to e-shopping and also stay alone. Hence, they have the most need for them. Security risk, privacy risk, time-loss risk, delivery risk, physical risk had insignificant effect on e-shopping BI among these older segments.
6.4. Multi-Group Analysis
Analyzing the PLS-MGA (Table 5
), considering the significant factors in SEM analysis, it could be noted that, there were not many group differences between early and later LC stages.
While performance risk was expected to be higher among later LC stages than early, results were to the contrary. Early LC stages were found to perceive a higher level of performance risk than the later LC stages; FN(AD) and FN(AI) were particularly found to perceive a low level of performance risk. Time loss risk, as expected, is one dominant factor that was found to affect families with young children. Families FN(I) showed a significantly higher time-loss risk than families without children (NM and EP). FN(I) was also found to differ with certain later LC stages, FN(TC) and EN&SS. In a similar vein, though not so rigorous, FN(PC) and FN(SC) perceived very high time-loss risk that it also differed with some later LC stages (EN&SS).
Contradicting expectation, the after-sale risk perceived by early LC stages was higher than the later LC stages—FN(AI) and EN&SS, that they differed significantly with all early LC stages except FN(I) groups. EP and FN(SC) perceived higher levels of after-sale service risk. As expected, perception of psychological risk was found to be much higher among later LC stages, especially FN(AI) and EN&SS, that it differed with all early LC stages. EN&SS differed with later LC stages, FN(TC) and FN(AD) as well, in that people belonging to this stage perceived a high level of psychological risk.
With regard to social risk, among later LC stages, FN(AD) perceived a higher level of social risk than any other group, that it differed with FN(SC) and FN(TC). EN&SS also perceived a higher level of social risk, but not as high as FN(AD). However, among other later LC stages, FN(AI) or FN(TC) did not differ significantly from the early LC stages. Hence their perceived social risk was small. Source risk was found to be perceived highly by EN&SS stages and it differed with EP and FN(SC) stages. FN(TC) and FN(AD) also differed with EP stages. However, FN(AI) did not show a significant difference with any early LC stage.