• Failure to involve beneficiaries and users: those who can ensure that project designs are well-matched to local realities.
• Rigidity in project delivery: following a pre-planned approach such as that mandated by methods like Structured Systems Analysis and Design Methodology, or narrow use of LogFrames.
• Failure to learn: not incorporating lessons from experience that arises either before or during the ICT4D project.
• Ignoring local institutional capacities: not making use of good local institutions where they already exist or not strengthening those which could form a viable support base.
• Ineffective project leadership: that is unable to direct and control the ICT4D project.

This does not represent an exhaustive list of causes but one can find one or more of them in many failed ICT4D projects.  And they are deliberately selected because – if we turn them around to their mirror-image project enablers – they become the five key components of the “process approach” to development projects: beneficiary participation; flexible and phased implementation; learning from experience; local institutional support; and sound project leadership.

The process approach arose during the 1980s and 1990s as a reaction to the top-down, “blueprint” approach[2].  The blueprint approach was particularly associated with use of foreign technologies in rural development projects.  Perhaps, then, it is no surprise that it has filtered through into ICT4D practice.

Equally, though, one can see elements of the process approach in action in successful ICT4D projects:

• Beneficiary participation: the M-PESA mobile finance project in Kenya incorporated the views of users into project design through user trials and volunteer focus groups.
• Flexible and phased implementation: India’s agricultural information kiosk project, e-Choupal, used a pilot approach for all new services; introducing them one-by-one and planning designs and scale-up on the basis of those pilots.
• Learning from experience: Grameen incorporated the lessons from its microfinance projects into the design and delivery of its Grameen Phone programme of rural mobile telephony.
• Local institutional support: Brazil’s community computing project, the Committee to Democratise Informatics, is founded on the development of local institutional capacity through each of the schools it creates.
• Sound project leadership: returning to M-PESA again, Vodafone put skilled project managers in place in Kenya in order to make the project work.

Each one of these projects – and one can no doubt find many others within the ICT4D field – demonstrates more than one of these five elements.  This is not unexpected since the process approach can be understood not as five rather arbitrarily-categorised, separate components but as an integrated whole.  It can be conceived like a wheel (see figure below[3]): flexible, phased implementation being the tyre that absorbs the bumps as the project goes along, feeding contextual information to learning from experience: the central axle from which the spokes of participation, local institutions and leadership radiate, giving strength to the whole.

 Figure 1: The ICT4D Process Approach Wheel

The process approach also reconceives the notion of success in ICT4D projects.  Instead of seeing either success or failure as cross-sectional, final judgements on a project, instead – like a point on the rolling wheel – any judgement must be seen as contingent and passing.  Instead of success and failure, we would therefore talk of multiple “successes” and “failures” as the project proceeds.  Any overall judgement would rest on relevance of the ICT4D solution, opportunities for capacity building, and sustainability.  A process approach contributes to each of these.

And for ICT4D practitioners, a process approach can help pose questions:

• What is the role of beneficiaries throughout the project’s stages?
• What is the mechanism for changing direction on the project when something unforeseen occurs?
• What is the basis for learning on the project?
• What local institutions can be used for project support?
• What is the nature of project leadership?

And so forth – these and other questions can lead to concrete plans, schedules and roles which incorporate the lessons of the process approach into future ICT4D activities.

This blog entry is a summary of the online working paper “Can a Process Approach Improve ICT4D Project Success?“, published in the University of Manchester’s Development Informatics series.

If you have experiences of ICT4D project failure or success to share, please do so via comments.

International curricula – such as those provided by organisations like the IEEE and the ACM – offer an off-the-shelf template for this updating.  But African universities are going to face challenges in implementing these curricula, which were designed for Western (typically US) rather than African realities.  And when curriculum change is introduced, African universities and Education Ministries need a systematic means to evaluate progress, to highlight both successes and shortcomings, and to prescribe future directions.

A recently-published case study – “Changing Computing Curricula in African Universities: Evaluating Progress and Challenges via Design-Reality Gap Analysis” – investigates these issues, selecting the case example of Ethiopian higher education.  In 2008, Ethiopia decided to adopt a new IEEE/ACM-inspired computing curriculum.  It moved from three-year to four-year degrees, introduced a new focus on skills acquisition, more formative assessment, greater diversity in teaching approaches, and a more practical engagement with the subject matter.

Most literature and most advice about changes to ICT-related curricula has tended to focus on content rather than process.  As a result, there has been a lack of systematic guidance around the implementation of curriculum change; particularly in relation to evaluation of change.

In the Ethiopian case, the design-reality gap model was brought into play since it has a track record of helping evaluate ICT-related projects in developing countries.  The explicit objectives and implicit expectations built into curriculum design were compared with the reality found after implementation.  This enabled assessment of the extent of success or failure of the change project, and also identification of those areas in which further change was required.

The gaps between design and reality were assessed along eight dimensions – summarised by the OPTIMISM acronym, and as shown in the figure below.

Using field visits to nine universities and interviews with 20 staff based around the OPTIMISM checklist, the evaluation process charted the extent to which the reality – some 18 months after the curriculum change guidance was issued by the Ministry of Education – matched the design objectives and expectations.

The evaluation found a significant variation among the different checklist dimensions, as shown in the figure below. 

For example, the new curriculum expected a combination of:

• Specialist computer classrooms to support advanced topics within the subject area, and
• General-purpose computer classrooms to teach computer use and standard office applications to the wider student body.

Yet in most universities, there were no specialist computing labs, and ICT-related degrees had to share relatively basic equipment with all other degree programmes.

Similarly, the spotlight focus of curriculum change on new student skills had tended to throw into shadow the new university staff skills that were an implicit design requirement for change to be effective.  The evaluated reality was one in which a largely dedicated and committed teaching community was hampered by the limitations of their own prior educational experience and a lack of computing qualifications and experience.

But progress in other areas had been much better.  The national-level environment (milieu) had changed to one conducive to curriculum change.  Formally, two new Educational Proclamations had been issued, supporting new teaching methods and new learning processes; and two new public agencies had been created to facilitate wider modernisation in university teaching.  Informally, Ministry of Education officials were fully behind the process of change.

Similarly, university management systems and structures had been able to change; assisted by the flexible approach to structures that was particularly found in Ethiopia’s new universities, and by a parallel programme of business process re-engineering within all universities.

Evaluation using the design-reality gap model was therefore a means of measuring progress, but it was also a means of identifying those gaps that continued to exist and which needed further action.  It thus, for example, led to recommendations of ring-fencing a capital fund for technology-related investments; some redirection of resources from undergraduate to postgraduate in order to deliver the necessary staffing infrastructure; and a reconsideration of some curriculum content to make it more Ethiopia-specific (in other words, changing the design to bring it closer to local realities).

There were challenges in using the design-reality gap model for evaluation of curriculum change: allocation of issues to particular OPTIMISM dimensions, and drawing out the objectives and expectations along all eight dimensions.  Overall, though, the model provided a systematic basis for evaluation, one that was assuredly comprehensive, and one through which findings could be readily summarised and communicated.

The full case study can be found here.  Other pointers are welcome to materials on computer science curriculum change in developing countries, including specific materials on the evaluation of such changes.

So here’s my question: “If e-government is so great at cutting costs, how come my taxes haven’t gone down?”

Of course, taxes depend on far more variables than just e-government.  But the simple answer to the question is “. . . because e-government does not save money, it costs money”.  That seems likely the case in the global North where e-government seeks to cut costs by replacing expensive humans with cheap technology.  It is most definitely going to be the case in the global South where the technology is more expensive and the humans are much cheaper.

Despite the obvious importance of the topic, good quality cost:benefit calculations are rare but can be found.  Six years of e-government in UK local government saw £3.90 billion of investment release just £0.97 billion of savings[2].  The aggregate cost:benefit ratio of e-government projects in Australia was 1.64:1[3].

Rarer still is good quality work from developing countries.[4]  However, a recently-published study of e-government in Bhutan by Mayumi Miyata provides a model for systematic and comprehensive evaluation of e-government costs and benefits. The case study focuses on the Road Safety and Transport Authority of Bhutan, which issues driving licences and vehicle registration documents.  This was traditionally a paper-based process, and often slow; particularly for driving licences which had to be sent by post from regional offices to the head office in Thimphu.  In the mid-2000s, an Internet-enabled database system was installed so the main information associated with these processes could be passed instantly between offices.  (This was therefore an “e-administration” application for use by government staff rather than an “e-services” application for use by citizens.)

Data for Miyata’s research was gathered both before and after the introduction of this e-government system including detailed observation and timing of work processes, a breakdown of departmental accounts, and a survey of citizens.  The “after” component was undertaken in 2007; two years after implementation of the system, allowing plenty of bedding-in time.

Activity-based costing showed that the direct labour cost for issuing licences and registrations fell 24% following introduction of e-government; from US$15,080 to US$11,530 per year[5].  For example, the direct cost of issuing one driving licence fell from US$1.57 to US$1.17.  This was achieved largely through a significant redesign and decentralisation of internal decision-making and workflow.

However, introduction of e-government brought additional costs – hardware, software, internet connectivity and the cost of IT staff – totalling US$11,080 per year (set-up costs being amortised over 10 years).  The only indirect saving was in reduced postal cost (US$720).  Thus, overall costs were US$15,800 per year before e-government; US$22,610 after e-government.  A rise of 43%.

We need to recognise some specific features of this case that make it typical of a least developed country:

• the particularly low labour costs and high IT costs;
• the relatively low volumes of transactions across which costs can be spread (the case is more akin to a local than national government in size);
• the use of e-administration rather than a web-based self-service system which, while still requiring human back-office intervention, would automate some processes.

Miyata’s research thus provides a model that should be replicated for a broader set of examples.

On the other hand, Miyata’s work misses out three additional reasons why e-government globally fails to deliver cost savings:

• the relatively high rate of e-gov project failure, the costs of which must be included in any overall cost:benefit accounting[6];
• the learning curve – often of some years – that must be traversed before e-government applications can be used efficiently and effectively[7];
• the need for government e-services to be run in parallel with existing face-to-face, phone and postal service channels in order to bridge the digital divide and avoid excluding large sections of the population from access to government services; public e-services thus being a supplement to, not substitute for, other channels[8].

Does this mean e-government is a waste of money, and we should ask for our US$3 trillion back?  No.  What it means is that e-government is not going to save money for government and help bring taxes down.  The benefits of e-government lie elsewhere.  Again, Miyata’s paper is a good illustration:

• External savings: the lead time from application to receipt was reduced by minutes, weeks, even months for outlying offices.  Wait times in offices may also have come down.  Other studies report shorter waits and fewer journeys.  Saving of journeys can be monetised, and saving of citizen time might be (it depends how that saving is spent).  The key cost savings of e-government may thus be external not internal: for service users not administrators.
• Internal control and accountability: e-government provided managers with greater oversight of work processes and staff.
• Service quality and equity: citizens reported the quality of service and the fairness of treatment improved after introduction of e-government.

Other research shows further qualitative and external benefits delivered by e-government including: greater transparency of public services; greater accountability of public servants and politicians; reduced corruption; lower costs for business; greater attraction of foreign investment[9].  Please comment to add your own examples of evidence.

So e-government may not bring your taxes down, but – if properly designed and implemented – it will bring a positive economic and social return on investment.

First question: what do you mean by “those who don’t have ICTs”?

We need something a bit more nuanced than a simple, binary digital divide, and can use instead a digital divide stack of four categories (see figure below):

- Non-Users: those who have no access to either ICTs or ICT-based information and services.

- Indirect Users: those who do not get hands-on themselves, but gain access to digital information and services via those who are direct users.

- Shared Users: those who do not own the technology, but who directly use ICT owned by someone else (a friend, workplace, ICT business, community, etc).

- Owner-Users: those who own and use the technology

Of course we would need to make transverse slices through the figure; potentially, one slice for each different type of ICT, but particularly noting many in developing countries would be in a different category level for mobiles compared to the Internet.

Second question: what’s the evidence on inequality?

It is relatively limited and often bad at differentiating which digital divide categories it’s talking about.  However, we can find three types of evidence.

The Rich Get Richer; The Poor Get Poorer: situations in which some category of user gains a benefit from ICT while non-users suffer a disbenefit.  For example, micro-producers of cloth in Nigeria who owned or had use of a mobile phone found they were gaining orders and income; micro-producers without mobile phone access found they were losing orders and income (to those who had phones). (See also work on growing costs of network exclusion.)

Development vs. Stasis: situations in which some category of user gains a benefit from ICT while non-users do not gain that benefit. For example, farmers in rural Peru who used a local telecentre were able to introduce improved agricultural practices and new crops, which increased their incomes.  Those who did not use the telecentre just continued farming in the same way as previously.

Spillover Benefits: situations in which some category of user gains a benefit from ICT while non-users also gain a (lesser) benefit.  One rather less-publicised outcome from the case of Keralan fishermen using mobile phones to check market prices is an example.  Those fishermen without mobile phones saw their profit rise by an average Rs.97 (c.US$2) per day as a result of the general improvements in market efficiency and reduced wastage which phones introduced.  This was about half the profit increase seen by phone owners and meant, even allowing for the additional costs, that returns to phone ownership were greater than those for non-ownership.  However, it was a spillover benefit to non-ICT-users.

ICT4D research on spillovers to non-users specifically has been rare, with the main interests in non-users being to understand why they are non-users; and most spillover work being done between sectors or enterprises and/or focusing on the spillover of encouraging ICT adoption rather than more immediate benefits.

This does seem to be changing, perhaps because of the growth of mobile and related to earlier work on the externalities to non-users of arrival of rural telecommunications.  Rob Jensen’s Kerala study found a second digital spillover: while fishermen’s revenues rose, the price per kg fell due to the increase in supply arising from less waste.  Fish consumers (many likely non-users) now paid less than previously thanks to the mobile-induced efficiency gains.  More directly, a study of M-PESA’s community effects in Kenya found its use providing positive financial, employment, security and capital accumulation externalities that affected both users and non-users within the community.

We also have a little evidence of spillover benefits from owner-users to indirect users:

- Follow-up work with Keralan fishermen found fish workers who will only get into a boat with a mobile phone-owner due to safety concerns, with these indirect users able to benefit from the owner should the boat get into difficulties.  That paper’s author (personal email) also gives the example of an indirect user citing as a benefit being informed of – and able to curtail – his daughter’s illicit elopement via his boat owner’s phone.

- Research on farmers in Northern Ghana[1] found those who did not themselves own or use mobiles benefitting from information passed on from phone owners, including more frequent meetings with agricultural extension officers; meetings that were coordinated by phone owners.

In all these cases, owner-users are benefitting more than the lower-category users to whom benefits spill over.  That means – if you’ll forgive the pun – that in these cases ICTs are causing all boats to rise but the ICT-using boats to rise somewhat faster.  Inequality may still grow; perhaps absolutely but not relatively.

I look forward to what appears to be forthcoming work by the Global Impact Study on non-user spillovers.  However, this remains a poorly-understood and little-researched issue; one that needs a greater focus since it is central to understanding the digital divide and digital inequalities.  It also has implications for practice; suggesting ICT4D projects should promote non-user spillovers as much as they promote ICT usage.  As ever, your pointers to spillover research and practice are welcome.

 Follow @CDIManchester

Actor-network theory began as a means to explain how science works, such as the operation of scientific laboratories and projects.  However, it has subsequently grown to be seen as a full-blown social theory.  In particular, ANT says three things.

First, it says, “Hey, sociologists, you’ve been so obsessed with humans that you’ve been ignoring all the objects in the world.  But those objects – documents, mobile phones, plants, websites, etc – play an important role; just like humans they shape the people and other objects around them. So ANT is going to treat them the same as people, and call them both ‘actors’.”

Second, it says, “Hey, sociologists, because you’ve been so obsessed with humans, you think that society and social contexts or social factors are what explains everything in life.  But you’re wrong.  In fact you’re so wrong you’ve got your basic equation of life the wrong way around.  You think that society explains what goes on in the world.  Nope.  What goes on in the world is what explains society.  So ANT is going to focus on the mechanics of life: the ways in which people and objects interact with each other.”

Third, it says, “Hey, more recent French-type sociologists, you’ve been so obsessed with breaking things apart to understand the bits of grammar and bits of history that made them that your idea of researching a clock would be to smash it to pieces with a hammer.  That is not how to research a clock.  To research a clock you need to understand how all the pieces got put together, following the network of people and objects that interacted in order to make that clock.  So ANT is going to focus on how networks are assembled.”

Much ANT writing is horribly obscure, so full of hideously complex sentences and words that the writers must surely have done this deliberately in the hope of avoiding Oscar Wilde’s dictum, “to be intelligible is to be found out”.  But, done well, ANT can tell a good story and even occasionally give you the sense that you are suddenly seeing the world in a whole new light.  A whole new light that – because it’s about dynamics and innovations and technology and networks – seems especially relevant to ICT4D.

A couple of good entry points – good because they each provide a fairly clear and portable conceptual framework that you can re-use in your own research – are:

-         Callon, M. (1986) Some elements of a sociology of translation: domestication of scallops and the fishermen of St Brieuc Bay, in: Power, Action and Belief, J. Law (ed.), Routledge & Kegan Paul, London, 196-233

-         Law, J. & Callon, M. (1992) The life and death of an aircraft: a network analysis of technical change, in: W.E. Bijker & J. Law (eds), Shaping Technology/Building Society, MIT Press, Cambridge, MA, 21-52

Also not too unreadable is Latour’s Reassembling the Social, though had Latour been shot half-way through the dialogue with a PhD student that is reported in the book, I can’t help feeling a verdict of justifiable homicide would have been returned.

Although, as noted, use of ANT in ICT4D research has been limited there have been enough examples, at least from developing country cases within the information systems field, that we get a sense of the questions ANT is good at answering:

-         How do you explain the trajectory of an ICT4D project?

-         What role does technology play in an ICT4D project?

-         How does power manifest itself in an ICT4D project?  How were apparently powerless actors able to influence the direction of an ICT4D project?  How was it that apparently powerful actors didn’t get their way on an ICT4D project?

-         How does a particular ICT4D innovation (be it a new technology or business model or idea) diffuse or scale-up or sink without trace?

-         How did a particular ICT4D impact or ICT4D policy come about?

If you’ve identified other ICT4D questions that are especially suitable for an ANT lens, then do contribute them.

If you want an example of applying ANT in ICT4D that also includes a reflection on the pros and cons of the theory, and some thoughts on applying it in your research, I can recommend:

-         Stanforth, C. (2007) Using actor-network theory to analyze e-government implementation in developing countries, Information Technology and International Development, 3(3), 35-60

There is also a discussion of the relation between ICT4D and ANT in:

-         Rubinoff, D.D. (2008) Towards an ICT4D geometry of empowerment: using actor-network theory to understand and improve ICT4D, in: Developing Successful ICT Strategies, M.H. Rahman (ed.), Information Science Reference, Hershey, PA, 133-154

And feel free to comment on other ICT4D literature that makes use of ANT.

If you would like to participate in discussions about ANT, you can join our online forum on LinkedIn at: http://www.linkedin.com/groups/ActorNetwork-Theory-in-Development-Studies-3995328

We will also be populating a group on Mendeley with reference details, and welcome contributions: http://www.mendeley.com/groups/1255941/actor-network-theory-in-development-studies/

In the 1980s, Robert Solow triggered the idea of a productivity paradox, saying “You can see the computer age everywhere but in the productivity statistics.”  And for many years there was a similar developing country growth paradox: that you could increasingly see ICTs in developing countries except in the economic growth data.

That is still largely true of computers and to some extent the Internet, but much less true overall as mobiles have become the dominant form of ICTs in development.  In particular key studies such as those by Waverman et al (2005), Lee et al (2009), and Qiang (2009) have demonstrated a clear connection between mobiles and economic growth and/or between telecoms more generally and economic growth.  They all address the “endogeneity” problem: that a correlation between telecoms (indeed, all ICTs) and economic growth is readily demonstrable; but that you then have to tease out the direction of causality: economic growth of course causes increased levels of ICTs in a country (we buy more tech as we get richer); you need to try to control for that, and separate out the interesting bit: the extent to which the technology causes economic growth. 

The studies try to do this and show ICT investments cause economic growth, but they are all multi-country and provide no specific insights into the experiences of a particular developing nation.  If you know of such data, do please contribute.  Meanwhile, a recent edition of “Kenya Economic Update” provides an example.  Some overall points:

• The ICT sector grew at an average of nearly 20% per year from 1999-2009 (by contrast, Kenya’s largest economic sector – agriculture – shrank by an annual average of nearly 2% per year).
• The number of phone subscriptions has grown from the equivalent of one per 1,000 adults in 1999 to the equivalent of nearly one per adult in 2010; Internet usage rates for 2010 were around four per ten adults.
• Person-to-person mobile money transactions at the end of 2010 were equivalent to around 20% of GDP with two of every three Kenyan adults being users.

But the report’s strongest claim is this: “ICT has been the main driver of Kenya’s economic growth over the last decade. … Since 2000, Kenya’s economy grew at an average of 3.7 percent. Without ICT, growth would have been a lackluster 2.8 percent—similar to the populaton growth rate—and income per capita would have stagnated”.  So ICTs were responsible for 0.9 of the 3.7% annual GDP growth, and for all of Kenya’s GDP per capita growth.  Put another way, ICTs were responsible for roughly one-quarter of Kenya’s GDP growth during the first decade of the 21^st century.

Other nuggets from the report and from original World Bank data underlying the report:

• The “ICT sector” is actually the “posts and telecommunications” sector.  Comparing figures from Research ICT Africa for mobile + fixed line + Internet/data services with those for the overall sector suggests that ICTs form by far the majority (likely greater than 90%) of that sector.  For the ICT part of the sector, latest figures for 08/09 show mobile takes a 54.8% share, fixed line takes 39.5%, with 1.8% for Internet services and 3.8% for data services (not 100% due to rounding).
• The ICT sector in 2009 still represented only 5% of total Kenyan GDP (compared to 21% for agriculture/forestry), and growth has been volatile, at least as based on the recorded figures, ranging from 3.5% per year up to 66% per year during the first part of the decade, and from 7.9% to over 30% during the second part of the decade.  Only tourism (hotels/restaurants) was more volatile.  In six of the ten years of the 2000-2009 decade, though, ICT was Kenya’s fastest growing sector.
• In the first half of the decade, annual investments in mobile were higher than annual revenues; but the ratio has subsequently slipped to investment averaging around half of revenue.  Investments in mobile during 2001/02 to 2009/10 are estimated at US$3.2bn (c.KSh250bn) and US$3bn in fixed phone services, with broadband, Internet and BPO investments adding perhaps another US$1bn.
• The ICT sector provided a more than six-times-greater contribution to Kenyan GDP in 2009 compared to 1999.  Directly, the ICT sector contributed to 14% of the country’s GDP growth between 2000 and 2009 (at constant (i.e. not actual/current but accounting for inflation) prices, it grew from KSh13.7bn in 2000 to KSh71.8bn in 2009; GDP overall grew from KSh976bn to KSh1.382tn).  So the World Bank’s calculation that ICTs contributed a quarter of GDP growth during the decade also include a specific, quantified assumption about ICTs triggering growth in other sectors, in particular the financial sector.
• Employment in the ICT sector is estimated to be around 100,000 in 2011 (c. 0.7% of the estimated 14m overall labour force).  But ICT punches above its weight in other ways: changes in mobile prices at the start of 2011 were credited with both causing the Kenyan inflation rate to drop and with potentially derailing government constitutional talks due to the substantial knock-on effects in causing tax revenues to drop since phone companies now contribute such a significant proportion of government income.

So, overall, what do we have here?  Some fairly solid evidence that ICT sector growth (predominantly due to mobiles) is making an important direct contribution to economic growth in this developing country.  And some less clear evidence that the indirect GDP growth effect of ICTs may nearly double this.  Thanks to mobile money, Kenya has seen a particularly strong take-up and economic role for ICTs, but it is fairly typical in terms of mobile investment, revenues, subscriber base, employment, etc.  In that case, it’s not too much of an extrapolation to expect that ICTs will have contributed something like one quarter of GDP growth in many developing countries during the first decade of the 21^st century.  Evidence of ICT impact that development strategists and practitioners should be more aware of.

The case is one of “socially-responsible outsourcing” (SRO): the use of – in this case IT – outsourcing contracts to drive livelihood benefits directly into poor communities.  I’ve already written up an impact analysis of SRO that delivered new jobs, incomes, skills and empowerment into poor urban communities in Kerala.

The case study analysed here also involved SRO to poor communities in India, but this time to telecentres in rural Bihar.  It contracted from city-based clients both data work (data entry, data editing, digitisation for three clients) and voice work (call centre-based service/technical support and tele-sales for six clients including the Govt of Bihar).  The full case report is available online.

The Development 2.0 promise is that it will bypass traditional development blockages to bring digital production – that is ICT-based productive work – to the bottom of the pyramid.  If that was the promise, what was the reality of this pilot project, run by the social enterprise Drishtee?

The first reality is that this is far from ‘direct development’.  It is a re-intermediated model of development that interposes two layers between urban clients and village production: a city-based head office that interacts direct with clients, a regional office based in a large village (6,000 inhabitants) in rural Bihar which can undertake both data and voice work and quality assurance of the third layer: individual telecentres in relatively remote villages where data (but not voice) work can be done.

The second reality is that the technical and human infrastructure in rural areas requires significant investments before it can get close to the promise of this type of Development 2.0.  The regional office (20 PCs, two printers, 512 kbps Internet connection via VSAT and ISDN, UPS and generator: see Figure 1) had to be created at a cost of US$13,000.  The village telecentres (at least two PCs, GPRS Internet link (114 kbps),and electricity plus back-up: see Figure 2) were within 35km of the regional office and were already in existence.  They had cost an average US$1,500 to set up with running costs (inc. loan costs, rent, telecoms, maintenance) of US$150 per month.  Some needed additional investment to ensure greater reliability of power supply.

Figure 1: Main rural outsourcing office (Source: Drishtee)

Figure 2: Village telecentre (Source: Drishtee)

The staff who were to do the work in both the large and the remote village locations were selected from unemployed youth (presumed to be under 25 years old) who had some school education including English language skills and IT familiarity.  However, they all required two-three month training programmes covering IT, language, typing, and communication skills before there were seen as ready to participate in this particular part of the digital economy.  Even then, their initial accuracy rate for data work was around 75%, rising to 95% after about two months of work.  They still required the layered superstructure of quality control between them and the clients.

In all, the pilot project created 19 new jobs in the large village (regional office) and 5 overall in the village telecentres, with earnings of US$80 per month (for 25 days of eight-hour shifts; a pay level set at the top of the typical US$40-80 range for rural business process outsourcing work) when there was sufficient work.  In such circumstances, the telecentre owners could net US$90 per month from the SRO, thus strengthening telecentre sustainability.  In addition to the creation of jobs and incomes at the bottom of the pyramid, this project confirmed the findings of the Kerala SRO programme that there are key gains in skills and self-confidence.

If the message is that the BoP isn’t quite ready, but can be made ready, for this particular fraction of Development 2.0, the news from the top of pyramid is less cheery.  Having largely addressed the technical, skill and quality challenges of SRO, Drishtee’s main difficulty has been demand: getting enough clients.

They charge US$1 (Rs.45) per job hour for domestic clients, which is the going rate, and rural outsourcing has clear advantages over outsourcing to urban areas (c.35% cost advantage, and much lower staff turnover rates than the c.40% per year in urban locations).  But there have been difficulties of awareness of the rural/socially-responsible outsourcing model, and of trust of the model and of a new entrant into the field like Drishtee.

Scaling – even sustaining – this particular model is therefore difficult.  Experiences in Kerala show that both scalability and sustainability are achievable, but those all occurred within one large state government rather than via the more commercial sales and marketing approach that Drishtee must follow.

In conclusion, the Bihar pilot demonstrates that the benefits of the digital economy – specifically, ICT-based jobs – can be brought to rural areas, and can deliver livelihood benefits of income, skills, and empowerment.  The poor in rural communities therefore do not just have to be digital consumers, they can also be digital producers.

It is also an example of ICT helping bring new development actors into play; in this case a multi-layered social enterprise that provides a new form of intermediation between urban business and rural livelihoods.  It is disappointing that the same constraints we got bored of discussing in the 1980s – power, telecommunications, skills – are so deeply persistent.  And troubling that new constraints – trust, awareness, demand – may be holding back realisation of Development 2.0′s potential.  But increasing numbers of new intermediaries are bringing ICT-based SRO to poor urban and rural communities, so we can expect that realisation to increase in future.

Links: see also blog entry on BoPsourcing: Fighting or Fuelling Inequality?

Follow @CDIManchester

I’m going to nominate W. Howard Gammon writing in Public Administration Review in 1954.  Please comment with earlier nominations, but otherwise, W. Howard Gammon becomes the godfather of e-government research.

Of course Gammon’s review article: “The Automatic Handling of Office Paper Work” doesn’t mention e-government: according to Heeks & Bailur’s “Analyzing e-Government Research”, “The term ‘electronic government’ seems to have first come to prominence when used in the 1993 U.S. National Performance Review, whereas ‘e-government’ seems to have first come to prominence in 1997.”

However, Gammon is writing about the use of ICTs in the public sector: which is a common definition of e-government.  Hence, his is an article about e-government, even though computing was just in its infancy with, as he notes, some technical literature available but very little written for a management audience and nothing – until his review article – for a public management audience.

In some ways things were very different then.  Even by around 1990, there were more than 1 million computers in use across the US federal government.  Back in 1954, there were roughly forty computers installed in total, half “large-scale” such as the UNIVAC I (weight 13 tonnes, c.2,000 operations per second, memory <1kb; cost c.US$10m in today’s terms) and half punch-card-based “baby computers” such as the IBM-604 (c.100 cards per minute, program of up to 40 steps, monthly rental cost c.US$5,000 in today’s terms plus a shift team of 2-10 supervisors and operators).  Most were in the Department of Defense with a few in the Atomic Energy Commission, Census Bureau and Bureau of Standards.  There was a pilot application to automate selection of optimum procurement bids, and plans to apply computers for use in air traffic control, taxation and weather forecasting.  These applications were part of a broader expenditure (in 1952) of more than US$1.5billion (c.US$12billlion in today’s terms) on “adding, accounting and other business machines” within US public and private sectors combined; by 2008, total spending on ICTs in the US was roughly US$1.2trillion annually – a one hundred-fold increase in spending on ICTs.

However, the more striking thing that echoes across the decades is not how different but how similar the issues in the 1950s were to those we still face today.  The following examples illustrate:

a) Skill Set: E-Gov Needs Systems Skills More Than Technical Skills: “…it is not necessary to know how to make, or even to repair, these machines in order to make use of them.  For the public administrator … the emphasis needs to be placed on how and when to use these new devices” (p63).  Just so, for those learning today about e-government, understanding technical aspects is of relatively limited importance; much more important is to understand the application of the technology.  Put another way, e-government must be approached from an information systems not an information technology perspective: “it is a systems job which depends more on knowledge of what must be done, and why, than on knowledge of what makes electronic computers tick.” (p73).

b) Skill Set: E-Gov Needs Hybrids: a socio-technical approach is required that combines understanding of the ‘business’ of government with knowledge about the application of technology.  Such a combination could be undertaken within a team: a “joint effort between the business managers and the engineers, so that engineers may learn enough about the businessman’s problem to translate the requirements of the job into machine procedure and so that management staff may learn enough about the capabilities and limitations of electronic machines to allow management staff to visualize how the new devices can be applied and how the … organization must be changed to take full advantage of the capabilities of the new equipment.” (p67)  Such a combination might also be effected within a single person to create a socio-technical “hybrid” individual.  But in that case, it will be far easier to hybridise a mainstream manager than an IT person: “As the Metropolitan Life Insurance Company found in its study, it is far easier to teach company management specialists what they need to know about the possibilities and limitations of electronic data processing than it is to teach electronic engineers about the internal operating problems of the life insurance business.” (p73).  The exact same findings were reported in the 21^st Century for e-government in Chapter 12 of Heeks’ book “Implementing and Managing eGovernment”.

c) Implementation: E-Gov Needs Re-Engineering Not Just Automation: More than thirty-five years before Hammer’s exhortations to stop “paving the cowpaths” and stop “automating the mess”, Gammon had already identified the limited gains to be made from automation, and the need to start improvements by re-engineering the business processes of govenrment: “One quick generalization may be made: the introduction of an electronic information processing system is not like buying a new adding machine which can be plugged in as part of an existing established clerical routine. It would be foolish and wasteful to make the large investment required to install electronic methods without first conducting a careful study which begins with considering the basic objective of the operation” (p73) … “The effective application of electronic methods in a given organization requires a rethinking of its organization and procedures. When electronic methods are applied, many of the intermediate reports and steps in the transmission of information become unnecessary and should be eliminated.” (p72)

d) Implementation: E-Gov Needs Top Management Support: “Rapid progress can be made during such an investigation only if the management representatives are high enough in the organization to make the broad decisions regarding the methods of operation” (p67).  In the same way, more recently, top management support is still identified as a key necessity in successful e-government projects and its absence as a key cause of e-government project failure.

e) Implementation: Politics Matters in E-Gov: “there are organizational, procedural, economic, and social problems which must be resolved before automatic operation of … an office can be realized.” (p63).  Some of these problems relate to internal politics given the danger that ICTs in government will cause “the disturbance of established bureaucratic empires” (p72), thus making political factors an important cause of e-government failure.  This further explains why top management support is needed in implementation of e-government: “It also requires a broad point of view which looks to the good of the organization as a whole without being too much concerned about the effects of changes in methods on particular vested interests in the agency.” (p73).

f) Impact: E-Gov Affects Clerical Not Professional and Managerial Jobs: for lower-level clerical jobs, ICT brought the threat of “lowered prestige, relative decrease in real income, threat of unemployment, and routinization of many office skills” (p66).  That has come to pass: around 25% of federal white-collar employees were in clerical/typing work in 1952.  By the mid-2000s that had fallen to roughly 7%, largely as a result of new technology[1].  Meanwhile, skilled professionals would either be upskilled: “our accountants will then be free to do the more important job of analyzing and interpreting financial reports for management.” (p66) or unaffected: “there is no real possibility that the executive or the top administrator will become obsolete as the result of foreseeable advances in the use of electronic equipment.” (p73).  And there were already signs that shortages of ICT professionals would slow the rate of e-government: “the shortage of qualified experts to design, build, program, and service these electronic data processing systems will keep this possible revolution from taking place rapidly.” (p67).  More than fifty years later, Heeks (2006:101) still writes “The dearth of competencies is a major brake on the spread of e-government”.

g) Impact: E-Gov Impact Assessment Fails to Account for Total Cost of Ownership: there have always been ambitious claims for ICT in government e.g. that it “can make substantial savings and render better service” (p63).  But on the savings side, e-government impact calculations often focus just on cost savings (e.g. of labour) but fail to include the costs of ICT.  When the latter are taken into account, overall gains can disappear.  Gammon’s paper is suggestive of this: he reports a case where preparation time for monthly reports fell from forty person-days to six/eight hours.  Given clerical costs (in 1954 prices) of around US$200 per month, that would represent savings of about US$400 per month.  Yet according to P.B. Hansen in “Classic Operating Systems” the IBM Card Programmed Calculator on which this saving was achieved cost US$1,800 per month in rental to which would have to be added the costs of calculator operations staff.  Government’s tendency for lavish spending on ICTs was also already in evidence with reference to a Navy-organised symposium on “moderately-priced” computers; that criterion being defined as those costing (in today’s terms) less than US$1million.

Gammon’s paper is as much a review of then-current ideas about computing, drawn largely from the private sector, for a public sector audience as it is about computing in the public sector.  However, this focus means it still stands eligible for recognition as the first e-government journal article.

How, overall, should we read it?  I invite you to choose from its reflecting:

- “La plus ça change, la plus c’est la même chose”

- The failure of e-government practitioners to take note of key lessons known right from the start of IT in the public sector, given the continuing absence in e-government projects of many of the skill and implementation factors identified all those years ago.

- The failure of e-government researchers to find much new to say: you can see these same issues still in the conclusions of many of today’s e-gov journal articles.

Click here to link to a blog entry on the first application of e-government in a developing country.

What and Who: this is an informal (i.e. unregistered) micro-enterprise run by Goran, who is a student living in a major city in Eastern Europe.  He has been playing in-game himself to make gold and he also paid two other people, whom he knows via personal connections, to play characters on his account.  He has taught them how to play.

Making Virtual Gold: because WoW players have their characters on a specific European-based server, they can only buy gold that has been produced by a gold farmer on that server.  Goran has therefore placed his gold-farming accounts on two (out of a possible c.250) servers he thinks are particularly “high-yield”, i.e. active and popular with players.  His high-level characters have particular in-game “professions” – abilities – that are used to make or gather items (e.g. gathering herbs using high-level Herbalism and then making items with high-level Alchemy, or making items with Inscription).  These items can then be sold in-game, typically via the game’s auction system, thus producing in-game gold.  (He has chosen this method as more productive than what he sees as the approach used by many Chinese gold farms of fighting non-player characters, picking up items dropped when the NPCs are killed, and then selling these for gold.)

Selling Virtual Gold: most of the virtual gold has been sold to merchant-portals – the Web sites through which the great majority of players buy gold.  The two key portals used in this case are www.pusada.com and www.offgamers.com (OGM).  Players who wish to obtain in-game gold will buy from these merchants via their Web portals.

OGM are Malaysia-based and the sales process is undertaken via chat on their portal or via mobile phone calls.  When a customer indicates through the Web site that they are ready to buy, Goran is sent the in-game name of their character by OGM, and OGM calls the customer on their mobile.  Goran and the customer log into the game and the gold is transferred from Goran’s to the customer’s character: Goran sends screenshots to OGM as confirmation for his payment.  If Goran is late, OGM will call him on his mobile, so this is not quite the anonymous process it might appear.  OGM outsource some parts of their sales work (e.g. customer chat) to China.  Likewise Pusada – who are US-based – outsource to sub-contractors in other countries.  Unlike OGM, Pusada headquarters staff never go in-game but only broker the sale via IM chat.

Speaking English is thus a key part of Goran’s added value.  His sub-contractors do not speak English and so, for example, are unable to sell the items they produce in the WoW auction houses to make the in-game currency; Goran has to do that himself.  Likewise, thanks to his English, Goran is able to make money in a different way – by acting as the broker for a local gold farming workshop, selling their gold to the portals above and also to www.net4seller.com and (now defunct) www.mmoinn.com.  He charges the local workshop a commission of about 12% for this service.

Financials: Revenues are affected by two main forces.  First, the price that merchants charge to the final consumers.  This has been steadily falling as virtual gold continually devalues against real-world currencies.  For example, from March 2009 to January 2010 the US buy price of 1,000 WoW gold depreciated from US$14.50 to US$7.00.  Second, merchants price and pay in US dollars (even though the virtual gold is sold in euros to the European player-buyers).  During periods when the dollar appreciates against Goran’s national currency, this further erodes revenue.  From March 2009 to January 2010, the US$ rose by around 10%.  The result was a fall in Goran’s revenue from US$7.10 per 1,000 WoW gold in March 2009 to US$3.37 in January 2010.

Costs are made up of three elements: the cost of the World of Warcraft game accounts; game cards which cover the cost of a certain amount of in-game play; and the wages paid to his ‘sub-contractors’.

Nominally, Goran’s payments to those who farm the gold for him are paid on the basis of the amount of time they spend in-game with his characters rather than, for example, on the basis of what they produce in terms of either items or gold.  (He tried the item-based approach but it became too complex due to the variety of items produced and different prices earned; pay-per-gold is not possible since the players themselves are unable to sell their items in-game for gold.)  He paid just over US$7.00 per eight-hour working day, which was almost exactly the same as the minimum wage for a young person in this country.  However, as noted, this is nominal – over time, the pay would be more akin to a certain percentage of the overall profits made during a month.

The figures can also be used to get a sense of the mark-up being charged by the merchant-portals.  In March 2009, Goran was paid an average of US$7.10 per 1,000 WoW gold, and the average sale price from portals was around US$14.50: a roughly 100% mark-up.  This was very similar in January 2010, when Goran was paid an average of US$3.37 per 1,000 WoW gold by the portals, while their sell price was around US$7.00.

Balance Sheet

The balance sheet below shows a sample nine-day period in January 2010 during which ten sales were made: all but one via the merchant-portals.  Below the revenues earned from these sales, Goran’s calculations for total number of hours his sub-contractors spent in-game are shown.  Hours are converted to days and paid at just over US$7 per day with account and game card expenditure then added.  The balance is shown at the end, indicating a daily profit for Goran just greater than the national minimum wage.

Postscript: By March 2010, Goran had expanded his gold farming business to employ nine farming sub-contractors and a business manager to oversee them all, helping to pay his way through university.  By March 2011, though, Goran has – at least for the moment – stepped out of the gold farming business; in part to focus on his studies, but also because profitability is increasingly difficult.  At the time of writing, the advertised buy price on www.offgamers.com was US$1.00 per 1,000 WoW gold (i.e. less than one-third what it had been a little over a year previously); the advertised sell price was US$1.54.  This means the sub-contractor model would no longer work well.  Goran has been making some money by renting his WoW game accounts and a GatherBuddy bot (for automating in-game activity: like gold farming, this is also against game rules) license on a monthly basis to his former sub-contractor who was doing some gold farming himself, though finding it ever-harder to make a profit and, at the time of writing, likely to drop out because prices were so low.  Goran is currently offering his full support to US calls for revaluation of the renminbi!

One explanation comes from the concept of “institutional dualism”.

First developed to explain how Japan reacted to the import of Western ideas and technologies in the 19^th century, this can also be used to understand any innovation – including ICT4D initiatives – in which there is some separation between designers and intended adopters.

Each of those two groups sits within its own institutional network: a complex of institutional elements (e.g. norms, rules, beliefs, values) and organisations and actions.  Left to its own devices, any institutional network will tend to “self-reproduce”.  For example, its cultural values will encourage particular actions, and those actions will in turn reinforce the network’s cultural values.

But innovations like an ICT4D application will bring two different institutional networks – those of the application designers, and those of the application adopters – into contact.  We can call this institutional dualism because of the two institutional networks that come into play (see figure).

A strong example of institutional dualism would occur if a team from a European university designed an ICT4D application and then introduced it into a rural location in Africa.  The European designers’ behaviour is enabled by a set of Western organisations and shaped by a set of Western institutional forces, some of which will be inscribed into the ICT4D application.  During implementation this network is drawn into contact with the very different network of rural Africa, with different organisational structures, behaviours, and institutional forces.

Many ICT4D projects will be a bit less starkly drawn than this, but will still involve institutional dualism because designers and adopters almost always come from different places and different spaces.

What then happens?  There are four possible outcomes from a situation of institutional dualism:

• Domination: one of the institutional networks prevails over the other in the ICT4D project.  If the designers dominate, the project could fail due to its mismatch to the broader local context.  If the adopters prevail, that requires a complete re-design of the project to have occurred.
• Contest: neither of the institutional networks prevails, but there is ongoing competition between them.  The ICT4D project may stagger on, but always in difficulty as it is pulled in two different directions.
• Parallel-Running: a separation is arranged with some aspects of the project guided by the designers’ institutional network, some by the adopters’.  This is only possible where the ICT4D project has a broad flexibility and scope.
• Hybridisation: the ICT4D project becomes the site within which the two institutional networks blend, forming a mixture of institutional values and hence a set of hybrid actions within the organisational structure of the project itself.

Of these four, only hybridisation and some types of domination are likely to lead to a sustainable ICT4D project.  We have seen this in practice with a large-scale ICT4D case study of institutional dualism from the Brazilian public sector.  Although giving some outward signs of hybridisation, beneath the surface this remained a story of ongoing contest and parallel-running even some years after its first implementation.  It was still contingent, and it demonstrates the great difficulty ICT4D projects have in institutionalising themselves when operating in environments of strong institutional dualism.