---

---

GENDER DIFFERENCES IN THE PERCEPTION AND USE OF E-MAIL
IN TWO SOUTH AFRICAN ORGANISATIONS

 

Jean-Paul W. G. D. Van Belle

Adrie Stander

University of Cape Town

 

Abstract. The effect of user acceptance on system use has been researched extensively in the past and many models have been developed. The Technology Acceptance Model (TAM) is one such model that is based on the theory of reasoned actions in the information systems field. An extension to this model, researched by Gefen and Straub in 1997 includes the effect of gender on the model variables. They found that women perceive e-mail as more useful but more difficult to use than men do. The use of e-mail was found to be the same, regardless of gender. This study replicates the original study in two South African organizations: “older” knowledge workers in a commercial organization and younger students at our University. Our findings differ, depending on which sub-samples are considered. Overall, the difference which Gefen and Straub found in perceived ease-of-use and perceived social presence of e-mail appears no longer valid. These results could be explained through the use of non-parametric statistics or the specific demographics of the sample. Overall, however, there is a significant difference in perceived usefulness and, consequently, actual usage rates of e-mail. Our findings therefore, although not consistent with Gefen and Straub’s, seem to validate most of the TAM.

 

Introduction

 

The Technology Acceptance Model (TAM) has been widely used to explore attitudes towards technology and the resultant effect on use. This study attempts to explain the relationship between technology and gender in a South African context, using the TAM as the methodological approach.

 

It has been established that there is discrimination against women in technology-related fields and it is possible that electronic communication can eliminate or reduce certain gender issues in the future, due to the lack of physical interaction (Venkatesh and Morris, 2000). Gefen and Straub (1997) conducted a study entitled “Gender Differences in the Perception and Use of E-mail: An Extension to the Technology Acceptance Model”. This study took place in North America, Europe and Asia using knowledge workers from a number of airlines. This study is a replication of the original study, using a South African sampling frame.

 

Technology Acceptance and Gender

 

The Technology Acceptance Model is based on the theory of reasoned actions, applied specifically to the information technology field. Davis’ theory (1989) combines user perceptions of Usefulness (U) and Ease of Use (EOU) to determine user behavior. His grounding for this basis includes self-efficacy theory, the cost-benefit paradigm, research on adoption of innovations and the channel disposition model.

 

The model states that while U is influenced by EOU, both U and EOU predict the user’s evaluation of the desirability of using the system. This is referred to as Attitude towards use (A). Combining A and U determine the user’s Intention to use the system (I). Finally, actual use of the system is derived from I (Matheison, 1991).

 

Straub (1994) proposed the addition of an exogenous “social presence” or “information richness” factor to the TAM. Gibson et al., 1997). In 1997, Gefen and Straub extended Straub’s research model to evaluate The basis for this study was derived from an analysis on cultural differences among countries (gender effects on the TAM variables.

 

Christie (1997) and Bennett and Brunner (1998), found that the feminine attitude towards technology looks right through the machine to its social function, while the masculine view is more focused on the machine itself. E-mail is an attractive technology for women as it allows them to connect, empathize and share feelings, which is the basis of female communication

 

Coates (1993) suggests that both language and gender are developed through our participation in everyday social practice and that there is considerable evidence that the patterns of interaction typical of all-female groups differ from those of all-male groups.

 

Gender issues continue to affect business in technology-related fields. Copas (1996) says that, through technology, gender will be stripped away and the criteria will move towards who has the credentials and who can communicate, and keep and deliver commitments.

 

Research Methodology

 

The objective of this research was to examine the effect of gender on technology acceptance by the application of Straub’s extended TAM model (1994) and to compare it to Gefen and Straub’s study (1997). Two samples were drawn: one of mature (“older”) knowledge workers in a commercial organization and one of younger students at our University. Where reference to both groups is made, this is denoted by “South African”. It is recognized (as did Gefen and Straub) that these groups are unlikely to be representative of the entire population of South African e-mail users, but then it is hard to visualize exactly what would constitute a manageable representative sample.

 

The four hypotheses measured by the study are as follows:

 

H1: South African women will perceive the social presence of e-mail to be different (higher) than men (the c² statistical analysis will be able to test for difference but not direction, e.g. greater/higher or smaller/lower).

 

H2: SA women will rate the perceived usefulness of e-mail different (higher) than men.

 

H3: SA women’s rating of perceived ease of use of e-mail will be different (higher) than that of men.

 

H4: Given the truth of H1 through H3, SA women’s use of e-mail will be different (greater) than that of men.

 

The questionnaires used for the study were based on the questionnaire used by Gefen and Straub in their study. Reported e-mail use was measured by the number of e-mail messages sent and received. The perceived use and usefulness scales were taken from Davis’ original study (1989). Extensive pre-testing was performed on these scales, including a cluster analysis, and they have been found to be accurate, reliable and valid in other studies (Mathieson 1991; Adams et al, 1992; Straub, 1994; Chin and Todd, 1995). Social presence was measured using items taken from Short et al. (1976).

 

Two samples were chosen to give a more accurate representation of the population. The sample groups were:

 

Knowledge workers of SANS Fibres, a manufacturing company that produces yarn out of synthetic fibres. SANS Fibres was chosen due to the fact that a large number of the employees are knowledge workers and therefore represent a suitable sample group of mature users for the survey.

 

Senior university students. Second-year university students registered for an Economics II course at the University of Cape Town were selected because of their ICT skills and variety of backgrounds, including Commerce, Social Science, Engineering and Arts. They are therefore a good representation of the student population and, to perhaps a lesser extent, of a new generation of ICT users.

 

A total of 616 questionnaires were distributed in total. Of these, 296 were returned and found usable. Not all respondents answered all questions, so smaller numbers were found in some tables.

 

Data was gathered from knowledge workers in a company and second-year students, which limits the generality of the findings. It is possible, for example, that there is a bias present in the industry assessed, or in the University faculty of the students that responded.

 

The adoption of Straub’s conceptualization of social presence (SP) combined with the information richness (IR) of the medium (SPIR) also introduces limitations to the findings. This is due to the fact that the research paper focused on the social presence factor and IR was not specifically measured. It has been suggested by socio-linguistics that women have a higher sense of SP and that is why the authors focused on SP in this empirical study. (Gefen and Straub, 1997)

 

Consistent with Straub’s methodology, each of the questions had answers on a 7-point Lickert scale, ranging from “disagree strongly” to “agree strongly”. Each question was then related to one of the conceptualizations (variables): perceived usefulness, ease-of-use and social presence. To obtain a score for each of these three variables, the average of all the questions for the variable was calculated.

 

Data Analysis

 

The age and gender profiles for the two samples are given in the following tables.

 

Table 1. Age profile of SANS Fibres respondents

 

	16-20	21-30	31-40	41-50	51+	Unknown
Female	0	7	11	16	2	0
Male	0	15	36	35	19	2

 

Table 2: Age profile of Economics students

 

	16-20	21-30	31-40	41-50	51+	Unknown
Female	51	22	0	0	0	1
Male	39	39	1	0	0	0

 

These tables show that the average age of the students is indeed much lower – by more than twenty years – than that of the SANS workers. The samples therefore represent two different generations.

 

Figure 1 presents a summarized view of all responses per sampled group for each of the variables. Note that each variable was measured by a number of questions.

 

 

Figure 1. Summary of response by group and variable

 

 

Visual inspection suggests the following observations:

 

Females at SANS perceive a much greater usefulness and ease of use than males at SANS.

 

Ratings at SANS are almost without fail higher (i.e. the respondents agree more strongly) than those of the corresponding UCT group. This is especially so for the perceived usefulness.

 

Gender differences in the SANS sample are much more pronounced than in the UCT sample.

 

Gender differences in the SANS sample are in the opposite direction of those in the UCT sample.

 

The analysis to follow will show that not all of the above observations are statistically significant.

Before proceeding with the statistical testing of the hypotheses, it was necessary to check the distribution of the data. A c² goodness-of-fit test confirmed that the data was not normally distributed. For most of the study, therefore, a non-parametric statistical test, namely the c², had to be used. This is in contrast to the Gefen and Straub study where the large sample size resulted in much more normal distributions and hence the use of parametric statistical tests was possible.

 

Findings

 

 

The c² test was used to determine if any association existed between gender and the TAM variables: perceived usefulness, perceived ease of use and social presence. The following results were obtained (refer to the addendum for a sample calculation).

 

Table 3: Association of gender and the TAM factors for SANS Fibres

 

SANS	Perceived Usefulness	Perceived ease of use	Social Presence
D2	14.282	10.02	5.71
Degrees of Freedom	3	3	4
Significance level	0.25%	1.84%	>10%

 

D² refers to the calculated test statistic i.e. the sum of the squared differences (between actual and expected frequencies) for each cell. D² is then compared against the critical c² value for the corresponding degrees of freedom, i.e. (number of rows – 1) x (number of columns – 1). Where the significance level exceeds 5%, the difference can be attributed to randomness. The higher the value of D², the smaller the probability that this is resulting from chance.

 

Note that the degrees of freedom are usually less than 6 because columns had to be combined in order to obtain a sufficient expected frequency in each cell. D² refers to the calculated test statistic i.e. the sum of the squared differences (between actual and expected frequencies) for each cell. From the above data the authors found that female knowledge workers of SANS Fibres perceive e-mail to be more useful than their male counterparts at a 99% level of confidence. The differences in perceived ease of use are also statistically significant, though only at the 95% level of confidence. However, it was found that there is no significant difference in gender perceptions of the social presence of the medium.

 

Table 4: Association of gender and the TAM factors for Economics II students

 

UCT	Perceived Usefulness	Perceived ease of use	Social Presence
D2	12.96	4.00	8.34
Degrees of Freedom	5	3	3
Significance Level	2.38%	>10%	3.95%

 

In the case of the students the authors found significant differences in gender perceptions of the perceived usefulness and social presence of the medium, but no significant differences for the easy of use. A possible explanation may be the technology because, unlike the SANS employees, UCT students have a personal choice of mail clients, including web browsers. It is quite possible that female students have favored different mail client software than their male counterparts. The authors prefer to attribute this to the fact that the younger generation has grown up with ICT and its familiarity has reduced interface or usability problems. Unfortunately this could not be confirmed by the data.

 

Table 5: Overall results of the tests of association between gender and the TAM factors

 

Combined	Perceived Usefulness	Perceived ease of use	Social Presence
D2	17.70	1.82	2.86
Degrees of Freedom	5	3	5
Significance Level	0.34%	>10%	>10%

 

The results of testing hypotheses H1 through H3, making use of the combined data from both samples (due to the fact that that gender differences in the two samples are in opposite directions, the differences in the combined samples are relatively smaller) and the Chi-squared association test, were as follows:

 

The first hypothesis (H1) could not be strongly confirmed in the study, due to the fact that  there was no significant association between gender and social presence for the combined groups and SANS workers. However, there was a statistically significant association for the student population.

 

It was found that there was a significant association between gender and perceived usefulness (H2) for both the knowledge workers at SANS Fibres, the UCT students and the combined sample.

 

 

Although the SANS workers showed a significant association, no significant association was found between gender and perceived ease of use for students or the combined groups, therefore the third hypothesis (H3) was not strongly confirmed, either.

 

Comparison with Previous Study and TAM

 

Gefen and Straub’s original study conducted in North America, Asia and Europe had the following findings:

 

Figure 2. Diagrammatic representation of Gefen and Straub’s results

The diagram shows that the betas for H1 and H2 were found to be highly significant at the 5% level. As mentioned, Gefen and Straub used parametric statistics, namely Partial Least Squares analysis. Their test statistics are the betas, representing regression coefficients with a “gender dummy variable”. Although these betas can be interpreted intuitively, they cannot be compared with the (non-parametric) D²  test values in figure 3 H3 was found to be highly significant, yet in the opposite direction than was hypothesized. The implication of rejecting H3 was that H4 was not confirmed. However, Gefen and Straub did find indirect effects on e-mail usage through the effects of gender on PU and SPIR, which both affect Use. Our study produced the following results:

 

Figure 3. Diagrammatic representation of the current study’s results.

 

It was found that overall there is only support for hypothesis H2 and not for H1 or H3. The interpretation of this in terms of the TAM seems to be as follows. “If the perceived usefulness of e-mail is perceived to be greater by females than males, and there is no perceived difference in the ease of use, females would be expected to make greater use of e-mail (H4).”

 

The reported usage is statistically significantly different for the two genders. The data (i.e. the reported number of e-mails received and sent) was not distributed normally and therefore a c² analysis was used again.

 

Each sample was divided into three subgroups, depending on the number of e-mails sent and received, with the cut-off values set such that the subgroups would be roughly of the same size. The classification of low/medium/high use is therefore a relative distinction, based on the sample group norm. It will be noted that the number of emails differs significantly between the SANS and student samples, with the former using email on average five times as much as the latter.

 

The SANS sample was divided as follows: £ 15 e-mails/day (“low usage”; n=46), between 15 and 30 e-mails/day (“medium usage”; n=47) and ³ 30 e-mails/day (“high usage”; n=50). Surprisingly, dividing the boundaries by a factor five produced a fairly well-spread distribution for the student sample: £3 e-mails (n = 60), between 3 and 6 e-mails (n = 32) and ³ 6 e-mails/day (n = 56). In both sub-samples, and thus for the sample as a whole, females used e-mail more frequently than males.

Table 6: E-mail use: number of respondent reporting e-mails sent/received per day.

 

SANS	( 15	16 – 29	( 30
Male	39%	31%	30%
Female	11%	39%	50%
UCT	( 3	4 or 5	( 6
Male	42%	29%	30%
Female	39%	14%	46%

 

Table 7 provides the D² statistics.

 

Table 7: Association of gender and reported use of e-mail .

 

	SANS	UCT	Combined
D2	10.27	6.32	9.29
Degrees of Freedom	2	2	2
Significance Level	0.59%	4.24%	0.99%

 

H4 holds at a significant level of confidence for both the sub-samples and the overall sample.

 

Conclusion

 

The study finds evidence for the hypothesis that that is a difference in the use and perception of e-mail between men and women in South Africa, but the differences depend on the sample i.e. generation of e-mail users

 

The results of this study do not fully support the findings of the previous study conducted in North America, Europe and Asia where it was found that women rate both the perceived usefulness and ease-of-use of e-mail to be different than men do. There are two possible reasons for this:

 

South African-specific factors and the age of the respondents; and/or

 

The fact that our sample size is smaller resulted in less-normal distributions. Hence we had to use a non-parametric test. It is more difficult to obtain statistical significance with non-parametric tests.

 

Table 8 summarizes the results of statistical differences found in this study and compares them to the Gefen and Straub findings.

 

Table 8: Statistically significant gender differences found  (5% level of confidence) .

 

Variable	Gefen and Straub	SANS	UCT	Combined (SA)
Perceived Social Presence	Yes	No	Yes	No
Perceived Usefulness	Yes	Yes	Yes	Yes
Perceived Ease of Use	Yes	Yes	No	No
Reported Use	Yes	Yes	Yes	Yes

 

Overall, the research validates most of the causal relations in the TAM very well. Although the perceived social presence does not seem to differ according to gender, women tend to value the usefulness significantly greater than the men. Younger women especially appear to have overcome the “technology” gap found by Gefen and Straub (whether this must be attributed to better training/education or to the improved usability of the e-mail client software cannot be ascertained) and there is no significant difference in perceived ease-of-use. Both combine naturally to produce a statistically significant higher actual use of e-mail by women.

 

References

 

Adams, D., Nelson, R. & Todd, P. (1992). Perceived usefulness, perceived ease of use, and user acceptance of information technology: A replication. MIS Quarterly 16(2), 227-247.

Bennett, D. & Brunner, C. (1998). Technology perceptions by gender. Education Digest, 63(6), 56.

Chin, W. & Todd, P. (1995). On the use, usefulness, and ease of use of structural equation modelling in MIS research: A note of caution, MIS Quarterly, 19(2), 237–246.

Christie, A. (1997). Using email within a classroom based on feminist pedagogy. Journal of Research on Computing in Education, 30(2), 146.

Coates, J. (1993). Women, Men and Language. New York: Longman Publishing.

Copas, K. (1996) Technology holds a key to women’s future success, Business First, 12(42), 30.

Davis, F.D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13(4), 318-339.

Gefen, D. & Straub, D.W. (1997). Gender difference in the perception and use of e-mail: An extension to the Technology Acceptance Model. MIS Quarterly, 21(4), 389-400.

Gibson, J. L., Ivancevich, J. M. & Donnelly, J. H. (1997). Organisations – Behaviour, Structure, Process. New York: McGraw-Hill.

Matheison, K. (1991). Predicting user intentions: Comparing the Technology Acceptance Model with the Theory of Planned Behaviour. Information Systems Research, 2(3), 173-191.

Short, J., Williams, E. & Christie, B. (1976). The Social Psychology of Telecommunications. London: Wiley.

Straub, D. W. (1994). The effect of culture on IT diffusion: Email and fax in Japan and the US. Information Systems Research, 5(1), 23-47.

Venkatesh, V. & Morris, M. G. (2000). Why don't men ever stop to ask for directions? Gender, social influence, and their role in technology acceptance and usage behavior. MIS Quarterly, 24(1), 115-139.

Addendum: Sample Calculation

The following table summarizes the (average) answer relating to the “perceived usefulness” for the SANS sample.

Actual	Disagree Strongly	Disagree	Disagree Somewhat	Neutral	Agree Somewhat	Agree	Agree Strongly	Total
male	0	1	5	12	40	31	18	107
female	1	0	0	2	7	9	17	36
Total	1	1	5	14	47	40	35	143

Visual inspection clearly reveals the difference in distribution. The median value for males is “Agree Somewhat”, whereas the distribution for females is heavily skewed, with its median at “Agree Strongly”.

In order to calculate the D 2 statistic, the first four data columns have to be grouped so that the expected frequency for each cell exceeds 5.

Actual	Disagree/ Neutral	Agree Somewhat	Agree	Agree Strongly	Total
male	18	40	31	18	107
female	3	7	9	17	36
Total	21	47	40	35	143

The expected frequencies, assuming independence between rows (gender) and columns (perceptions) are as follows.

The weighted squared differences show the contribution towards the overall D 2 statistic.

Variance	Disagree/ Neutral	Agree Somewhat	Agree	Agree Strongly	Total
male	0.33	0.66	0.04	2.56	3.60
female	0.99	1.97	0.11	7.61	10.69
Total	1.32	2.64	0.15	10.17	14.28

As can be seen, the major contribution comes from the relatively large proportion of females that tend to “agree strongly”.

The critical c ² value for 3 degrees of freedom at a 1% confidence level is 11.345, and at 0.5% it is 12.838. (The actual probability of this difference happening in the absence of correlation is actually 0.25%.) Hence the assumption that there is no correlation between gender and perception must be rejected.

Expected	Disagree/ Neutral	Agree Somewhat	Agree	Agree Strongly	Total
male	15.7	35.2	29.9	26.2	107
female	5.3	11.8	10.1	8.8	36
Total	21	47	40	35	143