The Social Impact Of Broadband Household Internet Access - CiteSeerX

CHIMERA WORKING PAPER NUMBER: 2005-06 CWP-2005-06-Social-Impact-BB.doc

The Social Impact Of Broadband Household Internet Access Chimera Working Paper Number: 2005-06

Dr Ben Anderson1 Dr Yoel Raban2 1

Chimera, University of Essex

2

ICTAF, University of Tel Aviv

Contact: [email protected] Abstract: Broadband changes everything. Or so we are told. But does it? There is only one way to find out – follow people who move from narrowband to broadband internet access and see what changes. This paper reports exactly this kind of analysis using data from a 2 wave European panel study (e-Living) to see if switching to broadband increases the time spent online, decreases the time spent watching TV, increases the amount of online spend and decreases the amount of other leisure activities. The results suggest, in the main, that switching to broadband made little difference for this group of early broadband adopters who were already heavy internet users and who already spent more online than those who did not switch to broadband. There is also no evidence of a TV or social leisure substitution effect.

http://www.essex.ac.uk/chimera/


Chimera The work reported in this paper is part of the scientific programme of Chimera, the Institute for Sociotechnical Innovation and Research at the University of Essex. Chimera is a post-disciplinary institute employing social scientists, computer scientists, engineers, anthropologists, psychologists, HCI practitioners and interface designers specialising in ‘socio-technical’ research and consulting. It was set up in April 2002 at Adastral Park, Suffolk as a research institute of the University of Essex. Chimera carries out research which combines the social and technological sciences to: •

generate insights into personal and social use of information and communication technologies,

•

ground technological innovation in an understanding of people,

•

provide analysis to support evidence-based 'information society' strategies and policies in the public and commercial domain.

We achieve this through a balanced programme of basic and applied research projects, consultancy and publication. For more information see www.essex.ac.uk/chimera Contacting Chimera Chimera Institute of Socio-Technical Innovation and Research Ross Building (PP1, ROS-IP) Adastral Park, Martlesham Heath, Ipswich, Suffolk, IP5 3RE UK

Tel: +44 (01473) 632238 Fax: +44 (01473) 614936 E-mail: [email protected] Web: http://www.essex.ac.uk/chimera/

Citing This Paper Readers wishing to cite this paper are asked to use the following form of words: Anderson, B., and Raban, Y. (2005) The Social Impact Of Broadband Household Internet Access. Chimera Working Paper 2005-06. Colchester: University of Essex. For an on-line version of this working paper and others in the series go to www.essex.ac.uk/chimera/publications.html © 2005, University of Essex All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted, in any form, or by any means, mechanical, photocopying, recording or otherwise, without the prior permission of the Director, Chimera.

© 2005, University of Essex http://www.essex.ac.uk/chimera/

Page 2 of 20


Table of Contents 1

The broadband ‘revolution’ ............................................................................................................................ 5

2

Relevant literature............................................................................................................................................ 5

3

Definitions ......................................................................................................................................................... 6

4

Penetration of Broadband Internet access in households ....................................................................... 7 4.1

Overall distributions of PC and Internet access modes ........................................................................... 7

5

Routes to Broadband ...................................................................................................................................... 8

6

The difference broadband makes ................................................................................................................. 8 6.1

Difference vs Causality ............................................................................................................................... 8

6.2

Internet time model ...................................................................................................................................11

6.3

TV time model ...........................................................................................................................................12

6.4

E-commerce model ...................................................................................................................................12

6.5

Social leisure model..................................................................................................................................14

7

Conclusions ....................................................................................................................................................14

8

Bibliography....................................................................................................................................................15


Page 3 of 20



Page 4 of 20


1

The broadband ‘revolution’

At the time of writing there is an almost ceaseless stream of news flashes and marketing messages trumpeting the future social and economic (not to ignore profitable) opportunities of broadband internet access in the home. This has not gone un-noticed in policy circles since the early 1990s (Bangemann 1994; DTI 1994). More recently and as a natural progression from the Bangemann report the eEurope 2005 Action 1 plan states: “broadband enabled communication, in combination with convergence, will bring social as well as economic benefits. It will contribute to e-inclusion, cohesion and cultural diversity. It offers the potential to improve and simplify the life of all Europeans and to change the way people interact, not just at work, but also with friends, family, community, and institutions.” ((CEC 2002) p8). But is this happening and what evidence do we have of the difference that broadband makes to the domestic user? This chapter uses data from the e-Living survey to assess at least some of these issues empirically.

2

Relevant literature

The processes of choosing a broadband connection, the characteristics of those who do so and the possible effects this has have been the focus of a number of studies (Hoag 1997; Madden and Simpson 1997; Anderson, Gale et al. 2002; Kridel, Rappoport et al. 2002; Lee, O'Keefe et al. 2003; Paynter and Chung 2003; Ida and Kuroda 2004; Robertson, Soopramanien et al. 2004) as well as many private and public market research surveys such as the Flash Eurobarometers funded by the European Commission (Gallup Europe 2002). A number of these papers (such as (Ida and Kuroda 2004)) concentrate on econometric approaches to modelling and forecasting demand based on price and charging schemes and as such are of less interest here. Indeed of necessity early research was based on choice experiments thus Madden and Simpson (1997) base their analysis of who would be likely to adopt broadband and what they would be prepared to pay on data from hypothetical statements. Of the others, Hoag used a single cross-sectional survey of cable modem and narrowband users in the USA to establish that there were few socio-demographic differences between the two groups (confirmed by Anderson et al (2003)) but that broadband users tended to make more use of FTP and the Web than narrowband users and also to spend more time on-line. She also showed that they made more use of a wider range of applications and were more satisfied with their Internet experience. Kridel and colleagues (2002) used a multinomial logit model to describe the broadband choice process of households in a cross-sectional survey conducted in the US in 2000. They found that age, income, education level, as well as price, are all highly significant predictors of choice in their model. Lee et al (2003) provide an analysis of the rapid take-up of broadband internet in South Korea between 1998 and 2001 by when up to 50% of all households had adopted. They focus primarily on supply side issues such as market competition, national policies and infrastructure investment. However they also note that the mobilisation of demand through IT literacy activities particularly targeting housewives, the elderly, military personnel and farmers may have had a significant effect in driving up demand and thus uptake. Lee et al also suggest that Asian cultures are more likely to use the Internet for inter-personal communication than non-Asian cultures and that the increased affordances for this aspect of usage may also have contributed to increased demand for broadband although, as we shall see, non-Asian cultures also make significant use (broadband) internet services and applications for social communication. Like Kridel et al, Paynter and Chung (Paynter and Chung 2003) used cross-sectional survey data from New Zealand and a factor analysis technique to uncover clusters of service values of narrowband and broadband users and then to use these loadings as the basis for the modelling of service satisfaction. They showed that cost is related to likelihood of future broadband usage and, as with Hoag, that early broadband adopters in New Zealand were socio-demographically distinct from narrowband users. Thus men were more likely to adopt than women as were those with higher computer skills and who used the internet more per day. Interestingly educational status, age and income made no difference. Robertson and colleagues (2004) use a household survey to analyse the factors affecting narrowband and broadband choice in the UK in early 2003. They found that educational attainment, disposable income and the presence of children were all indicators of internet adoption and had a marginally stronger effect for

1

http://europa.eu.int/information_society/eeurope/2005/index_en.htm © 2005, University of Essex


Page 5 of 20


broadband as opposed to narrowband at this time. Their results also suggest that downward shifts in broadband price will lead to many more narrowband users switching to broadband as the prices equalise and that this would be concentrated in higher income and ICT acceptance groups. Pew Internet data for the USA has been used to analyse the differences between broadband and narrowband users (Horrigan and Rainie 2002). This data suggested that broadband users spent more time online, did more things and did them more often than narrowband users. They also suggested that home broadband users were ‘typical early technology adopters’ being disproportionately well educated, wealthy and male. However, as we will elaborate below, such studies of difference tell us nothing at all about broadband related outcomes. Indeed recent authors have noted that few academic publications focus on the impact of broadband on social and personal issues in contrast to the developmental and macro-economic issues (Firth and Mellor 2005). Firth and Mellor note that the results of even these economic analysis tend to provide more rhetoric than empirical analysis and call for a diverse analysis of the outcomes of broadband internet access. The e-Living project which has followed individuals (or households) over time using a longitudinal panel method in order to distinguish between the effects of moving to broadband from the effects of heavy internet users becoming broadband users. We will provide an illustrative example of this key difference below. This chapter uses wave 1 (2001) and wave 2 (2002) of the e-Living longitudinal household panel survey to answer this question. After defining ‘broadband’ the chapter summarises the literature to date and then provides descriptive data on the uptake of household broadband internet access in the six countries of the eLiving survey (UK, Italy, Germany, Norway, Bulgaria, Israel). Rather than repeat previous analyses of uptake we use the unique longitudinal nature of the e-Living data to develop models of the ‘impact’ of switching to broadband between wave 1 (2001) and wave 2 (2002) on time spent using the internet, social leisure, TV watching and e-commerce. We focus on these issues because they are germane to a number of academic, policy and commercial pre-occupations: Does broadband internet access lead to more time being spent online? Or is it that people can do

-

more in the same amount of time that is available to them? Might broadband internet access lead to less social engagement in a dystopian future as some might

-

suggest (Nie and Hillygus 2002)? Might the usage of broadband either for content or for new forms of leisure lead to a reduction in TV

-

viewing and an attendant switch in potential advertising revenues? To what extent does broadband enable more (or more valuable) e-commerce transactions by

-

households? The chapter concludes with a summary of the results and a brief discussion of their implications.

3

Definitions

What do we mean by ‘broadband’? According to the ITU (ITU 2003) “Broadband is commonly used to describe recent Internet connections that are significantly faster than today’s dial-up technologies, but it is not a specific speed or service”. Recommendation I.113 of the ITU Standardization Sector defines broadband as a transmission capacity that is faster than primary rate ISDN, at 1.5 or 2.0 Mbit/s. Elsewhere, broadband is considered to correspond to transmission speeds equal to or greater than 256 kbit/s, and some operators even label basic rate ISDN (at 144 kbit/s) as a “type of broadband”. In this report, while not defining broadband specifically, 256 kbit/s is generally taken as the minimum speed and so we use the following definitions: •

Analogue = Narrowband Internet access (does not include ISDN)

•

Broadband = Cable modem/ADSL

We should note that in Wave 1 the e-Living survey item asking about Internet access mode had one merged category for ISDN, Cable modem and ADSL. As a result it was not possible to distinguish ISDN equipped © 2005, University of Essex http://www.essex.ac.uk/chimera/

Page 6 of 20


households from ADSL/cable modem equipped households in the wave 1 data. This was rectified at Wave 2. Where Wave 1 data are used and this distinction matters to the analysis, this problem is noted. However wave 1 data can be used for longitudinal analysis which examines those who had analogue modem access at wave 1 but had adopted broadband Internet at home by wave 2 (for example).

4

Penetration of Broadband Internet access in households

This section provides descriptive analysis of the take-up of Broadband Internet across the six surveyed countries using the Wave 2 (2002) data. For comparison we refer to the results of the Flash Eurobarometer [EB] 135 results collected at exactly the same time (November 2002). 4.1

Overall distributions of PC and Internet access modes

Table 1 compares the results of e-living wave 2 with Eurobarometer (EB) 135 (Gallup Europe 2002) and ITU (ITU 2003) data. In most cases the differences between the two surveys are in the order of +/- 4% which is to be expected. However the e-Living results for Germany estimate fewer Internet households in general and fewer ISDN/Broadband in particular than does the EB 135. Comparison with other sources such as the ITU’s Birth of Broadband Report which also contains mostly 2002 data suggests that whilst the e-Living figures for Germany may be an underestimate, the EB figures are certainly an overestimate. However the ITU data for Italy are higher than e-Living and the EB but lower for Israel and Norway. This highlights the risks of using small survey sample sizes to generate these kinds of analyses and it is not clear which of these data are the most reliable. Table 1: Broadband uptake in selected EU countries in 2002. % with Internet access

% of Internet households with analogue modems only

% of Internet households with at least ISDN

% of Internet households with broadband2

UK [Flash EB 135] 50.0% 85.0% 4.0% 12.0% UK [ITU BoB] 10.5% UK [e-Living wave 2] 47.0% 87.0% 3.2% 9.7% Italy [Flash EB 135] 36.0% 79.0% 10.0% 9.0% Italy [ITU BoB] 14.7% Italy [e-Living wave 2] 31.1% 77.3% 8.2% 13.0% Germany [Flash EB 135] 48.0% 56.0% 47.0% 34.0% Germany [ITU BoB] 21.6% Germany [e-Living wave 2] 41.5% 49.8% 39.8% 8.7% Norway [ITU BoB] 7.2% Norway [e-Living wave 2] 57.3% 30.0% 55.5% 15.0% Bulgaria [e-Living wave 2] 4.6% 90.9% 1.5% 7.6% Israel [ITU BoB] 14.1% Israel [e-Living wave 2] 43.5% 71.2% 6.1% 22.5% Table notes: Comparison of results from e-Living wave 2 and Flash Eurobarometer 135 both collected November 2002 (weighted for non-response) and the ITU’s Birth of Broadband 2002 data, rows may sum to more than 100% as multiple answers allowed The ITU figures are taken from the Annexe table on p21 of the report’s executive summary and it should be noted that there are apparent arithmetical errors within the ITU Report’s table. We have provided the corrected figures here. The ITU figures represent broadband subscribers as a % of internet subscribers (not internet households) and so may be larger than both the e-Living and EB ‘household level’ figures.

However the e-Living results suggest that in the e-Living countries, broadband Internet access was most prevalent in Israel and Norway with Italy and the UK roughly equal. Bulgaria may appear to have had broadband penetration rate similar to that of the other countries but this in fact represents five households! Germany may have been lagging slightly behind the UK and Italy but this requires confirmation as noted above.

2

Defined as ‘at least cable modem or ADSL’ – may also include other such as WIFI, Ethernet etc. © 2005, University of Essex


Page 7 of 20


5

Routes to Broadband

The two waves of e-Living data provide an excellent resource to analyse the routes that households are taking to broadband and, indeed, their routes out again. Table 2: Routes to broadband (e-living wave 1 and 2, 2002, all countries pooled, weighted for nonresponse)

Internet access 2002

Internet access 2001 No PC, no Internet

PC but no internet

PSTN

Broadband*

No PC, no internet

87.30%

11.10%

2.40%

1.60%

PC, no internet

6.00%

53.70%

4.20%

2.80%

PSTN

5.40%

24.50%

73.50%

10.80%

ISDN

0.60%

6.40%

7.40%

66.30%

Broadband

0.70%

3.60%

11.00%

15.80%

0.70%

0.70%

1.10%

Something else

Table 2 pools data for all six countries and suggests that about 0.7% of those who had no personal computer (PC) or internet access in 2001 had become broadband users in 2002, 3.6% of those with a PC but no internet had moved to broadband whilst 11% had a PC and narrowband internet access before switching. On the other hand 87% of those who had no PC or internet were still in the same state and 54% of those with a PC but no internet had also not ‘progressed’ to internet access of any kind. Intriguingly some 14% (1.6 + 2.8 + 10.8) of those who had some form of broadband (ISDN or above) in 2001 had reverted to narrowband or some even to no internet access by 2002. Clearly whilst there is adoption going on, there is also dis-adoption and considerable stasis (non-adoption) as has been reported elsewhere (Anderson 2004).

6

The difference broadband makes

We have noted that there is considerable excitement about the potential effects of broadband not only on internet use, e-commerce but also on other activities such as TV viewing and, in a negative frame, social and leisure activities. However there is little or no evidence for these ‘effects’ from anything other than crosssectional surveys which cannot, other than by error-prone respondent recall, account for the historical behaviour patterns of those who are broadband adopters. As a result the apparent difference in behaviour that are attributed to broadband may in fact be due to the previous internet history of those who are now broadband adopters. This confusion is apparently endemic in the ICT market research community at this time (e.g. (Nielsen//NetRatings 2003; Kerner 2004)) and also, more surprisingly, in the academic literature as well. As an example Nielsen//NetRating’s May 2003 press release on broadband implied that the difference in online time between narrow and broadband users measured in a cross-sectional survey was ‘caused’ by the switch to broadband. 6.1

Difference vs Causality

To continue with this example we might hypothesise that moving to broadband would increase the minutes spent online due to the probable switch to flat-rate pricing. On the other hand we might hypothesise that people would spend less time as the speed should enable them to complete their tasks more quickly. Or indeed there may be no difference in time spent online since there may be only a fixed portion of time available to individuals in any given day for internet use. Let us first consider the case of a cross-sectional survey.


Page 8 of 20


Figure 1: Mean minutes online per week for narrowband users and broadband users (e-living wave 2, 2002, weighted for non-response. Error bars = +/- 1.96*SE) Figure 1 shows the typical results of a cross-sectional survey and by comparing the means using the error bars we can see that in the UK, Norway and Israel broadband users do indeed spend more time online than narrowband users. However we cannot conclude from this chart alone that broadband is causing this difference, it is just as likely to be caused by differences in the kinds of people who are narrowband and broadband users.

Figure 2: Mean minutes online per week for narrowband users, broadband users and broadband adopters (e-living wave 1 and 2, 2002, weighted for non-response)


Page 9 of 20


Figure 2 addresses this problem by comparing the mean minutes spent online by narrowband users who did not move to broadband (Wave 1 or 2: PSTN 2001 & 2002) with those who did (Wave 1 or 2: PSTN -> BB (before/after)). Now we can see that those who moved to broadband were already heavier users of the internet before they switched, indeed this may have been one reason for them to do so. We can also see that it is only in the UK that moving to broadband from narrowband is associated with heavier internet use in terms of minutes per week online. But of course this is still a relatively simple picture. We do not know the influence of internet experience, age, gender or education, all of which may be mediators of internet use as may changes in lifestage or lifestyle such as getting (or losing) a job, becoming a parent or retiring. To do this we need to use multivariate techniques which aim to predict current (wave 2) behaviour on the basis of historical (wave 1) behaviour and changes between waves 1 and 2 such as adopting broadband or losing a job. Table 3: Longitudinal model specifications Wave 2 weekly Internet minutes Y

Dependent variables Wave 2 Wave 2 online weekly TV spending (euros) minutes in last three months Y Y

Y

Y

Wave 2 out of home leisure Y Y

Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y

Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y



Y Y

Y Y

Y

Y

Y

Y

Y

Y

Y

Y

Y

Independent Variables Country dummies (UK, Italy, Germany, Norway, Bulgaria, Israel) w1: minutes per week using Internet w1: minutes per week using TV w1: online spend in last 3 months w2: out of home leisure w1: gender (1=female) w1-w2: was single, now married w1-w2: was single, now living as a couple w1-2: got job w1-2: retired w1-2: became unemployed w1-w2: couple has split w2-w1: change in hours worked, non workers = 0 w1: Education level - GCSE or equivalent w1: Education level - A level or equivalent w1: Education level - Degree or equivalent w1: aged 16-24 w1: aged 25-34 w1: aged 35-44 w1: aged 45-54 w1: aged 55-64 w1: aged 65-74 w1: aged 75 - 84 w1: aged 85+ Broadband switch dummy: (1= Went from PSTN to BB w1 - w2; 0 = stayed with PSTN) w1: Internet experience (years) w2-w1: change in frequency of email sent to family and friends w2-w1: change in number of different online activities done in last 3 months (list of 10) w2-w1: change in minutes per day using TV w2-w1: change in minutes per week using internet

Table 3 shows the specification of all models. In the case of internet and TV time it is worth noting that in order to ensure an approximation to the normal distribution of the dependent variable we have used a log transform (square root). In each case we have pooled all internet users at waves 1 and 2 in all six countries in order to achieve a reasonable sample size but have included dummy variables for the countries as controls to try to negate inter-country differences.


Page 10 of 20


In the remainder of the chapter we present summaries of models we have developed to test the effects of adopting broadband on internet time, TV watching time and online spending as representative of the range of potential ‘impacts’ of broadband. The full regression results are available in Annex A. 6.2

Internet time model

In this case we want to know if moving from narrow to broadband is associated with an increase or decrease in time spent online whilst controlling for previous internet usage and other transitions. For the internet time model we use the time spent online at wave 1 to predict the time spent online at wave 2; a range of socio-demographic variables including a number of transition variables such as retiring or losing a job as well as dummy variables for each country. We also include a dummy variable for switching to broadband, wave one internet experience, a set of internet behaviour change variables and a measure of change in TV watching.

Figure 3: Internet time model results (only coefficients with p < 0.05 shown, values = beta [standardised coefficients], adjusted r sq = 0.349) Figure 3 shows the results of running this model. As we can see that the strongest predictor of current internet time by a considerable margin is the amount of time spent online last year. This is hardly surprising – people’s habits do not change that quickly. Switching to broadband is the next strongest predictor followed by changes in two ICT behaviours – the frequency of emailing family and friends and the number of different online activities. These last two suggest that there is at least a mutually reinforcing relationship between email and other activities and time spent online as we might expect. Thus the more email one sends and the more activities one does, the more time one spends online and it is interesting to note that the effect for © 2005, University of Essex http://www.essex.ac.uk/chimera/

Page 11 of 20


email is marginally stronger. Internet experience also has a positive effect as does the break-down in a relationship whilst being in a couple at wave one and being female have a negative effect. These suggest that in general, and controlling for the other effects, those in a couple and especially women do not use the internet as much as others. Recalling the full model specification (Table 3) we can see that retirement and losing a job make no difference and nor does gaining a job. Given that these events would all have an effect on time that could be given over to internet use the lack of an effect warrants further investigation. 6.3

TV time model

In this case we want to know if moving from narrow to broadband is associated with an increase or decrease in time spent watching TV whilst controlling for previous TV usage. As we can see the specification for this model is almost identical with the only difference being the use of TV watching time as estimated by the respondents at each wave as the dependent (wave two) and lagged (wave one) variables. We have also kept the longitudinal ICT usage variables to see if changes in online activity have any effect on TV time. This model proved somewhat ‘disappointing’. Although it explains 33% of the variance in the wave two TV watching time (adjusted r sq. = 0.33) the only two variables that proved statistically significant were the amount of time spent watching TV at wave one (beta = 0.52) and the educational level of the individual (for a degree or higher, beta = -0.11). No change in work or life situation nor a change either in internet access mode or online behaviour has any significant effect on time spent watching TV according to this data. 6.4

E-commerce model

In this case we want to know if moving from narrow to broadband is associated with an increase or decrease in overall online spending. This model is again broadly similar to the previous two but includes the wave one and two spend variables and leaves out the ICT behaviour variables relating to email as we have no a priori reason to suppose that this will be linked to online spending. However we include internet experience at wave one as it is known to correlate with online spending, the level of internet use at wave 1 (minutes online) and the change in time spent online between waves one and two.


Page 12 of 20


Figure 4: e-Commerce model results (only coefficients with p < 0.05 shown, values = beta [standardised coefficients], adjusted r sq = 0.208) This model is rather similar in form and results to that for online time except that now we find no effect for those who moved from narrow to broadband. Instead we can see that the amount of online spend in the previous year together with time online and internet experience are excellent predictors of current online spend. Doing more activities online is linked to spending more as is being married or partnered in the previous year, irrespective of current co-habitation status (measured by our non-significant transition variables). Gender (being female) has a negative effect suggesting that men spend more online than women. What we are seeing therefore is confirmation of our hypothesis that the ‘broadband effect’ discussed above is actually a confusion of differences between narrowband and broadband users at this stage in the uptake curve. We find no ‘broadband effect’ in this data. We find that the best predictors of the amount spent online are related to internet experience, usage and, perhaps, competence. Simply moving from narrow to broadband makes no difference. Interestingly we also have some evidence that certain groups of people (cohabiting couples) are likely to spend more online. Since our models do not (yet) contain other indicators such as working hours and status we can only hypothesise that this may be related to lifestyle/lifestage conditions such as dual-workers who make use of online shopping within the constraints of busy lives. However we would have expected in this case that the presence of children (not significant) would also show an effect and it does not.


Page 13 of 20


6.5

Social leisure model

In this case we want to know if moving from narrow to broadband is associated with an increase or decrease 3 in out of home social activity . This model is similar to the previous three but instead of the online spend variable we use an index of social leisure constructed form the following question items: 1. Frequency of partaking in leisure activities - play sport, keep fit or go walking 2. Frequency of partaking in leisure activities - go to the cinema, a concert, theatre or watch live sport 3. Frequency of partaking in leisure activities - have a meal in a restaurant or cafe, or go for a drink to a bar 4. Frequency of partaking in leisure activities - attend activity groups such as evening classes 5. Frequency of partaking in leisure activities - Meet with friends These items were measured on a scale of: 7

Most days

6

2-3 times a week

5

About once a week

4

About once a fortnight

3

About once a month

2

Several times a year

1

Less often

0

Never

The index is simply the sum of these scores for each item so the more frequently an individual takes part in these activities, the higher their score. As with television this model produces ‘disappointing’ results. Although it explains 45% of the variance, the only variables that showed any statistically significance effects on the wave 2 leisure score were the active leisure score for wave 1 and age (all age groups had significantly lower scores than the 16-24 year olds). Increasing work hours and getting a job were both marginal negative effects suggesting that with a larger sample size they prove significant. Getting broadband internet was not significant and nor was increasing time spent online. We have repeated this analysis for each component part of the social leisure index and the patterns are broadly similar with one exception. The number of different internet activities has a positive effect on frequency of meals out (item 3). Quite why this should be the case is unclear but it is possible albeit tenuous that increased social leisure opportunities are discovered through broader internet use.

7

Conclusions

In part this chapter is inevitably a history lesson. The penetration of broadband Internet was still at an early stage in 2001 and 2002. Israel ranked with the UK, Germany and Norway in terms of households with PCs but the UK had proportionately more narrowband (analogue modem) Internet households than any other country. ISDN dominated in Norway and, to some extent in Germany. Some households had multiple modes of access, the most frequent in the UK being analogue modem and Cable modem/ADSL and ISDN with analogue modem/Cable in Germany and Norway. This reinforces a finding from earlier research, which suggested that new cable modem, and ADSL users were retaining their analogue modem access in case of broadband service failure (Anderson, Gale et al. 2003). The four models that we have presented in this chapter have used unique longitudinal data to examine the effect of moving to broadband on four ‘representative’ activities of interest primarily to the ICT industry but also to policy makers – internet time, TV time, amount of money spent online and out of home social leisure activities. As we noted in the introduction to this chapter a range of authors from market research consultancies to academics have proclaimed the ‘impact’ of broadband on these aspects of life. It turns out 3

Unfortunately within this dataset we have no measures of intra-household social activity with which to test the hypothesis that heavier internet usage may lead to less intra-household communication. © 2005, University of Essex http://www.essex.ac.uk/chimera/

Page 14 of 20


that such ‘impacts’ are hard to substantiate when we have ‘before’ and ‘after’ data on the same individuals and can control for a range of other simultaneous life changes. We have seen that the greatest effect on time spent online is not moving to broadband, although this was significant, but the previous behaviour of an individual. We have seen also the strong positive effect on time online of the frequency of emailing friends and relatives thus confirming Kraut et al’s similar finding for the USA (2000) and highlighting that it is not only Asian internet users who are driven to a great extent by social communication. This should remind us that those looking for ‘killer apps’ may need to investigate social rather than consumption software more fully. Given previous findings using longitudinal data that getting household internet access had a negative effect on Television use (Gershuny 2003) the results we see here suggest that whilst this may be so, and may even have been a historical phenomenon, the result does not hold for switching from narrowband to broadband internet at home. Indeed our results demonstrate the resilience of time spent watching TV to a range of life transitions. Perhaps the result which will cause the most disbelief in the ICT sector is that we have no evidence that switching to broadband will have any effect on the amount of money individuals spend online. As with time online we can instead see a steady progression of online spend driven largely by experience not only in terms of years spent online but in terms of breadth of internet use. We can also see an overriding effect for successful previous e-commerce – since the driving factor for spend this year is the amount spent last year. In addition if the amount spent starts to fall the trend may be hard to reverse although with just two data points this is speculative. Changing the mode of access made no difference to the broadband adopters surveyed in 2001 and 2002. Of course this picture may have changed – those adopting broadband in 2002 were very early adopters and more recent adopters may exhibit different behavioural change. We must also be cautions because our sample sizes are relatively small and the number of switchers from narrow to broadband is low. We have only got 12 months of data so at most people have been using broadband for 12 months and any changes we see here may be a novelty effect. On the other hand it may be that significant behavioural change has yet to come about in these households. But we have argued, and hopefully demonstrated that we require longitudinal data to find out. Overall the effects we see are largely caused by experienced/heavy users moving to broadband and they were, at this time, an unrepresentative group of internet users. In the future and possibly quite soon in some countries as recent UK survey data suggests less experienced ‘average net users’ and indeed new users will move to broadband. In this case we may see more of an ‘effect’ although the constraints of everyday life suggest that it is only a small group of people in any cohort who exhibit significant behavioural change. There is simply not that much slack in most people’s lives for major shifts in behaviour in the short term. On the other hand we must also express some concerns about the potential benefits of broadband. We have see in the case of the internet time and online spend models that it is those who have the most experience and greatest breadth of use who are doing and spending more. If this pattern continues then broadband access will not change the structural problems already found in narrowband – those who have the knowledge and experience gain the most benefit whilst those who lack the skills, knowledge and perhaps self-confidence are left further behind as others have shown (Selwyn 2005). This is not an issue that will be solved by technology nor by policies that focus on penetration and access as opposed to utility, value and social outcomes.

8

Bibliography

Anderson, B. (2004). Passing By, Passing Through and Dropping Out. Chimera Working Paper. Colchester, UK, Chimera, University of Essex. Anderson, B., C. Gale, et al. (2003). Domesticating Broadband - what really matters to consumers. Broadband Applications and The Digital Home. J. Turnbull and S. Garrett. London, IEE: 153-176. Anderson, B., C. Gale, et al. (2002). "Domesticating broadband - what consumers really do with flat-rate, always-on and fast Internet access." BT Technology Journal 20(1): 103-114. Anderson, B., C. Gale, et al. (2003). Domesticating Broadband - What Really Matters to Consumers. Broadband Applications and The Digital Home. J. G. Turnbull, S. London, IEE: 153-176. Bangemann, M., Ed. (1994). Europe and the global information society: Recommendations to the European Council. Brussels, HLEG INFO-SOC. © 2005, University of Essex http://www.essex.ac.uk/chimera/

Page 15 of 20


CEC (2002). eEurope 2005: An information society for all - An Action Plan. Brussels, Belgium, Commission of the European Communities. DTI (1994). Creating the superhighways of the future: developing broadband communications in the UK. London, HMSO. Firth, L. and D. Mellor (2005). "Broadband: benefits and problems." Telecommunications Policy 29(2-3): 223236. Gallup Europe (2002). Flash Eurobarometer 135: Internet and the Public at Large - Results and Comments. Gershuny, J. (2003). "Web Use and Net Nerds: A Neo-functionalist Analysis of the Impact of Information Technology in the Home." Social Forces 82(1): 141-168. Hoag, A. (1997). Speed and the Internet: The effects of high speed access on household usage. 25th Annual Telecommunications Policy Research Conference, Alexandria, Virginia. Horrigan, J. B. and L. Rainie (2002). The Broadband Difference: How Online American’s Behavior Changes with High-speed Internet Connections at Home. Washington DC, Pew Internet & American Life Project. Ida, T. and T. Kuroda (2004). Discrete Choice Analysis of Demand for Broadband in Japan. Interfaces for Advanced Economic Analysis Discussion Papers. Kyoto, Kyoto University. ITU (2003). Birth of Broadband. International Telecommunication Union Internet Report. Geneva, ITU. Kerner, S. (2004). More Broadband Usage Means More Online Spending, ClickZ Network. 2005. Kraut, R., T. Mukhopadhyay, et al. (2000). "Information and communication: Alternative uses of the Internet in households." Information Systems Research 10: 287-303. Kridel, D., P. Rappoport, et al. (2002). "The Demand for High-Speed Access to the Internet." Topics in regulatory economics and policy 39: 11-22. Lee, H., R. M. O'Keefe, et al. (2003). "The Growth of Broadband and Electronic Commerce in in South Korea: Contributing Factors." The Information Society 19(1): 81-93. Madden, G. and M. Simpson (1997). "Residential broadband subscription demand: an econometric analysis of Australian choice experiment data." Applied Economics 29(8): 1073-1078. Nie, N. H. and D. S. Hillygus (2002). "The impact of internet use on sociability: time-diary findings." IT & Society 1(1): 1-20. Nielsen//NetRatings (2003). Broadband Revolutionizing Europe's Internet Behaviour. London, Nielsen//NetRatings. Paynter, J. and W. Chung (2003). "Factors influencing broadband uptake in New Zealand." Innovation: management, policy & practice 5(2-3): 170-188. Robertson, A., D. Soopramanien, et al. (2004). "Understanding residential Internet service adoption patterns in the UK." Telektronikk 4: 84-94. Selwyn, N. (2005). "Whose internet is it anyway? Exploring adults' (non)use of the internet in everyday life." European Journal of Communciation 20(1): 5-26.


Page 16 of 20


Annex A

Full regression model results

The following models were estimated using the xi: regress command in STATA SE version 9.0 to force the expansion of category variables (e.g. country, age group) into dummies. Where this has been done the contrast category is specified.

A.1

Internet Time Model

w1: internet minutes Country (Contrast = UK) Italy Germany Norway Bulgaria Israel Female Got married Became couple Got a job Retired Became unemployed Couple split Acquired children Qualifications (Contrast = none) GCSEs or equivalent (age 16) A levels of equivalent (age 18) First or higher degree Age (Contrast = 16-24) w1: aged 25-34 w1: aged 35-44 w1: aged 45-54 w1: aged 55-64 w1: aged 65-74 w1: aged 75 - 84 Change in work hours Moved from PSTN internet to broadband Internet experience w2-w1: change in frequency of email sent to family and friends w2-w1: change in number of different online activities done in last 3 months (list of 10) w2-w1: change in minutes per day using TV Constant

Coef. 0.009

Std. Err. 0.001

t 15.060

P>t 0.000

Beta 0.419

-0.166 -1.422 -3.890 2.637 0.774 -2.034 0.474 -1.197 2.856 1.052 0.697 3.049 -0.506

0.822 0.873 0.914 2.268 0.883 0.558 2.146 2.382 1.795 2.903 2.027 1.671 1.487

-0.200 -1.630 -4.250 1.160 0.880 -3.650 0.220 -0.500 1.590 0.360 0.340 1.820 -0.340

0.840 0.104 0.000 0.245 0.381 0.000 0.825 0.615 0.112 0.717 0.731 0.068 0.734

-0.007 -0.050 -0.132 0.031 0.029 -0.098 0.006 -0.013 0.041 0.010 0.009 0.048 -0.009

-1.539 -1.182 -1.849

1.349 1.342 1.375

-1.140 -0.880 -1.340

0.254 0.379 0.179

-0.071 -0.051 -0.083

-2.124 -1.702 -2.733 -2.685 -0.920 -3.391 -0.030 5.149 0.300 0.353

0.811 0.820 1.000 1.189 1.838 3.519 0.035 0.782 0.141 0.107

-2.620 -2.070 -2.730 -2.260 -0.500 -0.960 -0.850 6.580 2.120 3.310

0.009 0.038 0.006 0.024 0.617 0.335 0.394 0.000 0.034 0.001

-0.091 -0.074 -0.088 -0.070 -0.014 -0.025 -0.022 0.177 0.061 0.087

0.451

0.164

2.750

0.006

0.073

0.000 13.742

0.004 1.453

-0.090 9.460

0.926 0.000

-0.002 .

N Adjusted R-square

996 0.342


Page 17 of 20


A.2

TV Time Model Coef.

w1: tv minutes Country (Contrast = UK) Italy Germany Norway Bulgaria Israel Female Got married Became couple Got a job Retired Became unemployed Couple split Acquired children Qualifications (Contrast = none) GCSEs or equivalent (age 16) A levels of equivalent (age 18) First or higher degree Age (Contrast = 16-24) w1: aged 25-34 w1: aged 35-44 w1: aged 45-54 w1: aged 55-64 w1: aged 65-74 w1: aged 75 - 84 Change in work hours Moved from PSTN internet to broadband Internet experience w2-w1: change in frequency of email sent to family and friends w2-w1: change in number of different online activities done in last 3 months (list of 10) w2-w1: change in minutes per day using internet Constant

Std. Err.

t

P>t

Beta

0.022

0.001

20.500

0.000

0.542

-1.041 -0.810 -0.126 1.150 -0.270 -0.024 0.243 -1.152 -0.462 1.529 -0.380 0.615 0.356

0.277 0.294 0.307 0.760 0.292 0.186 0.720 0.799 0.602 0.973 0.679 0.561 0.499

-3.760 -2.750 -0.410 1.510 -0.930 -0.130 0.340 -1.440 -0.770 1.570 -0.560 1.100 0.710

0.000 0.006 0.682 0.131 0.355 0.899 0.735 0.150 0.443 0.116 0.576 0.273 0.476

-0.122 -0.084 -0.012 0.040 -0.029 -0.003 0.009 -0.037 -0.020 0.041 -0.014 0.028 0.019

-0.177 -0.770 -1.445

0.454 0.453 0.465

-0.390 -1.700 -3.100

0.697 0.089 0.002

-0.024 -0.098 -0.191

0.318 -0.089 -0.082 0.460 0.679 0.399 0.012 -0.416 -0.066

0.272 0.274 0.335 0.399 0.617 1.180 0.012 0.262 0.047

1.170 -0.330 -0.240 1.150 1.100 0.340 1.020 -1.590 -1.400

0.243 0.745 0.807 0.249 0.271 0.735 0.306 0.112 0.163

0.040 -0.011 -0.008 0.035 0.030 0.009 0.026 -0.042 -0.039

-0.030

0.036

-0.840

0.403

-0.022

-0.009 0.001 8.978

0.055 0.001 0.528

-0.160 0.700 17.000

0.870 0.487 0.000

-0.004 0.018 .

N Adjusted R-square

996 0.363


Page 18 of 20


A.3

e-Commerce Model Coef.

w1: online spend Country (Contrast = UK) Italy Germany Norway Bulgaria Israel Female Got married Became couple Got a job Retired Became unemployed Couple split Acquired children Qualifications (Contrast = none) GCSEs or equivalent (age 16) A levels of equivalent (age 18) First or higher degree Age (Contrast = 16-24) w1: aged 25-34 w1: aged 35-44 w1: aged 45-54 w1: aged 55-64 w1: aged 65-74 w1: aged 75 - 84 Change in work hours Moved from PSTN internet to broadband Internet experience w2-w1: change in number of different online activities done in last 3 months (list of 10) w2-w1: change in minutes per day using internet w1: internet minutes Constant

Std. Err.

t

P>t

Beta

0.001

0.000

7.360

0.000

0.212

-1.880 -0.673 -1.223 -2.317 -1.583 -0.252 0.242 0.141 0.438 -0.860 -0.436 -0.326 0.150

0.195 0.206 0.217 0.536 0.208 0.132 0.493 0.544 0.425 0.687 0.480 0.390 0.347

-9.650 -3.270 -5.640 -4.320 -7.590 -1.910 0.490 0.260 1.030 -1.250 -0.910 -0.840 0.430

0.000 0.001 0.000 0.000 0.000 0.056 0.624 0.795 0.303 0.211 0.364 0.404 0.665

-0.346 -0.110 -0.191 -0.125 -0.273 -0.056 0.014 0.007 0.029 -0.036 -0.026 -0.024 0.012

0.408 0.310 0.408

0.314 0.313 0.320

1.300 0.990 1.270

0.194 0.322 0.204

0.087 0.062 0.085

0.248 0.173 -0.292 -0.446 0.077 -0.568 0.013 0.104 0.131

0.191 0.193 0.236 0.281 0.426 0.833 0.008 0.188 0.033

1.290 0.890 -1.240 -1.590 0.180 -0.680 1.560 0.550 3.950

0.196 0.371 0.215 0.113 0.856 0.496 0.119 0.580 0.000

0.049 0.035 -0.043 -0.054 0.005 -0.020 0.044 0.017 0.123

0.092 0.002 0.001 1.369

0.039 0.001 0.000 0.342

2.370 2.200 4.050 4.000

0.018 0.028 0.000 0.000

0.068 0.072 0.141

N Adjusted R-square

. 1004 0.211


Page 19 of 20


A.4

Social leisure Model

w1: out of home social leisure Country (Contrast = UK) Italy Germany Norway Bulgaria Israel Female Got married Became couple Got a job Retired Became unemployed Couple split Acquired children Qualifications (Contrast = none) GCSEs or equivalent (age 16) A levels of equivalent (age 18) First or higher degree Age (Contrast = 16-24) w1: aged 25-34 w1: aged 35-44 w1: aged 45-54 w1: aged 55-64 w1: aged 65-74 w1: aged 75 - 84 Change in work hours Moved from PSTN internet to broadband w2-w1: change in number of different online activities done in last 3 months (list of 10) w2-w1: change in minutes per day using internet w1: internet minutes Constant

Coef. 0.587

Std. Err. 0.024

0.164 -0.602 -0.257 -0.612 -0.624 -0.327 0.351 -0.701 -1.479 -0.391 -0.997 -0.262 0.182

24.520

P>t 0.000

Beta 0.619

0.350 0.371 0.391 1.037 0.377 0.235 0.891 0.983 0.768 1.242 0.892 0.705 0.635

0.470 -1.620 -0.660 -0.590 -1.660 -1.390 0.390 -0.710 -1.930 -0.310 -1.120 -0.370 0.290

0.640 0.105 0.511 0.555 0.098 0.164 0.694 0.476 0.054 0.753 0.264 0.710 0.774

0.014 -0.046 -0.018 -0.014 -0.050 -0.034 0.009 -0.017 -0.046 -0.008 -0.027 -0.009 0.007

0.552 0.631 0.804

0.577 0.575 0.586

0.960 1.100 1.370

0.339 0.273 0.171

0.054 0.059 0.077

-1.077 -1.439 -1.602 -1.050 -2.651 0.393 -0.029 0.106 0.102

0.352 0.359 0.443 0.518 0.775 1.506 0.015 0.341 0.070

-3.060 -4.010 -3.620 -2.030 -3.420 0.260 -1.940 0.310 1.460

0.002 0.000 0.000 0.043 0.001 0.794 0.053 0.757 0.146

-0.099 -0.133 -0.109 -0.058 -0.087 0.006 -0.046 0.008 0.035

0.000 0.000 9.128

0.002 0.000 0.775

0.090 -0.940 11.770

0.931 0.346 0.000

0.002 -0.027 .

N Adjusted R-square

t

997 0.211


Page 20 of 20