Earlier this year, Africa held its first continent-wide conference on O.R. practice (ORPA-1) with partial funding support from INFORMS as well as EURO, IFORS, KPMG and other sponsors. ORPA-1 crowns five years of exciting developments, including the creation of two new Sub-Saharan African O.R. societies. This may signal a new era of O.R. activity in a region which had been the subject of many O.R. papers but which had not yet seen the impact of widespread application of O.R. methods.

We had the good fortune to be invited to present a plenary paper at ORPA-1, representing INFORMS. This article reports on ORPA-1, reviews broader developments in O.R. in Africa, and outlines a proposal for spreading O.R. across Africa via end-user modeling. In particular, we suggest that 800,000 end user modelers could be trained in Africa over a decade for about $6 million per year, a "dose" of O.R. that might make a material difference to the continent, and perhaps to our profession as well.

Report on the Conference


ORPA-1 was held in Ouagadougou, Burkina Faso, April 7-8. The conference drew 120 attendees from academia, industry and government. Most participants were from Burkina Faso, including many students from the University of Ouagadougou, but about one-third traveled from abroad. Under the leadership of General Chair Eric Soubeiga (KPMG, UK), ORPA-1 aimed to raise the profile of O.R. in Africa amongst the user-community and provide a forum for discussing O.R.'s potential to solve real-world problems. His Excellency Youma Zerbo, minister for technical and vocational training, presided over the conference opening, demonstrating local support for this mission.

Three plenary talks were given on "O.R. for Development in Africa" (Theo Stewart, University of Cape Town, for IFORS), "From decision theory to decision-aiding methodology: OR for a better world" (Alexis Tsoukiąs, Centre National de la Recherche Scientifique, for EURO) and our plenary on "O.R. & Public Policy: Past Successes and the Future Promise of End User Modeling." Stewart argued that in the developing world the emphasis has to be on improving effectiveness rather than efficiency, so O.R. should be applied within a framework of multiple-criteria decision analysis in a way that is transparent to stakeholders, allowing for wider participation in the policy making process.

Tsoukias traced the history of O.R., citing recent development in Africa as an important milestone. He also talked about both successful applications and challenges for the future, including topics relevant to Africa such as environmental management, governance and humanitarian security.

Figure 1: The number of countries covered by O.R. societies in Africa has grown substantially.

We argued that end-user modeling could be taught and deployed on a wide scale in Africa (800,000 end-user modelers) for a modest cost (about $6 million per year) to address development concerns (more on that below).

Papers were presented on topics ranging from supply chain management to an African perspective on the design of health care systems. There were tutorials - including one by Alistair Clark (University of the West of England, U.K.) on implementing O.R. with spreadsheets. There was also student participation, including an interesting paper by Tabitha Gathoni Mundia, a fourth-year undergraduate at University of Nairobi, Kenya. She explored "using minimal spanning theory to achieve the goal of having piped water in all homesteads around Kangaru village."

Much about the conference would have been familiar to INFORMS attendees, but there were twists, including the dominant language (French), dress (more colorful than a sea of business suits) and food (thanks to Blaise Some, University of Ouagadougou, who chaired the Local Organizing Committee). One provocative theme concerned the ethics of providing decision support to leaders in regimes plagued by widespread corruption, raising a new perspective on the perennial question of whether the O.R. analyst's role is to inform decision-makers or to recommend/advocate particular actions.

Table 1: Recent O.R. Conferences In Sub-Saharan Africa First African Conference on Operational Research (ACOR-1) - Ouagadougou, Burkina Faso January 2003 ACOR-1 gave rise to the West and Central African Network of Operations Research (WACANOR). The First International Conference in the Eastern Africa Region on Operations Research and Development (ORDA-1) Nairobi, Kenya September 2003 The conference was organized by the Operations Research Society of Eastern Africa (ORSEA). Its theme was the potential for Operations Research to play a role in African development. International Conference on Operational Research — Niamey, Niger January 2004 This conference, organized by the West and Central African Network of Operational Research (WACANOR), focused on using OR as a decision tool and included a tutorial/workshop sequence for practitioners without an OR background. Operational Research Practice in Africa (ORPA-1) — Ouagadougou, Burkina Faso April 2005. The ORPA conference, intended to be the first of a series, was organized in cooperation with INFORMS, IFORS, EURO, WACANOR and others.

Another discussion thread concerned how to get O.R. to decision-makers facing non-routine, time-critical missions. O.R. is routinely applied to long-term, non-routine decision-making (e.g., strategic planning) and to on-going, time-critical decision-making (e.g., production scheduling and inventory management), but crisis management and disaster response have a different character. O.R. quite literally can save lives in those applications, if processes can be established for rapid development and deployment of custom-tailored modeling.

An example serves to illustrate. David Smith presented a paper (De Angelis et al., forthcoming) on emergency food distribution by the United Nations World Food Programme. At one level this was an elementary transportation problem. Food arrived in three ports and was airlifted to 12 interior cities because the civil war made surface transport infeasible. However, there were complications: airports had limited capacity, planes could not park overnight where airport security was inadequate, there were pilot scheduling constraints, etc. These practicalities made the problem a moderately complicated routing and scheduling ILP problem. Not surprisingly, the optimal schedules could deliver substantially (about 20 percent) more food than could the judgmentally derived heuristic solutions actually employed in the field. So in principle O.R. could have saved many, many lives, but in reality it didn't, because the optimal solutions were produced ex post, after that crisis had passed.

There will be more such crises, and an O.R. expert could adapt the model to future exigencies, but the particular strategies that would have been optimal in Angola in 2001 are not likely to be optimal next time. What sort of intellectual infrastructure needs to be created to allow O.R. to serve these one-shot real-time applications? Should relief agencies hire expert O.R. analysts? If so, do O.R. education programs produce people well-rounded enough to function in the field? Should O.R. specialists take turns being "on call" the way doctors are, willing to drop what they are doing and respond to an emergency? If so, would relief managers know enough about O.R. to make the call and carry their end of the conversation? Can relief managers be trained as end-user modelers so they can build their own models and/or have more productive conversations with O.R. specialists?

The De Angelis et al. paper highlighted a third ORPA-1 theme that would be novel at INFORMS: the prominent role of non-governmental organizations (NGOs). INFORMS is familiar with O.R. informing private and government sector decision-makers, but the "interface" sector of non-profits, NGOs and civil society is increasingly important not only in Africa but also in our society (Drucker, 1992). Indeed, non-profits employ some 12.5 million people with revenues that reached $486 billion by 1998 (Lampkin and Pollack, 2002).

The Recent Spread of O.R. in Africa


ORPA-1 is just the latest in a series of exciting developments for O.R. in Africa that might be traced to the International Conference on Operational Research for Development (ICORD-IV) held in Kruger National Park in May, 2001 (Rand, 2004). Five years ago the only O.R. societies in Africa were those in Egypt and South Africa. Since then societies have formed in Tunisia, Algeria, Morocco, East Africa and West/Central Africa, so now a significant portion of Africa is covered by O.R. societies.

O.R. is by no means entirely new to Africa, but there has been something of a shift from O.R. done for Africa to O.R. being done by Africans in Africa. The new energy is not at all isolationist. There is great emphasis on international collaboration and networking. Africa seems, though, to be emerging as a source of O.R. analysts, not just as a source of challenging problems for non-African O.R. experts to model.

Table 2: Africa's share of English-Language O.R. & public policy articles compares favorably with whose of other regions Category Google Latin Scholar Latin JSTOR Latin Africa America Asia Africa America Asia Africa America Asia Physical 17% 6% 19% 16% 7% 18% 7% 4% 7% Health 49% 21% 42% 47% 23% 38% 3% 1% 1% Education 8% 3% 8% 18% 8% 19% 7% 1% 2% Economic 50% 20% 50% 35% 28% 44% 0% 0% 21% Public 54% 26% 44% 53% 25% 42% 17% 13% 13% Equity 71% 37% 66% 73% 43% 66% 50% 50% 50%

The traditional "O.R. for Africa" approach has been productive in terms of publications. A JSTOR search of top-flight social science and business journals yielded more than 4,000 hits for O.R. publications dealing with Africa. Indeed, in our area of expertise (O.R. & Public Policy), Africa's share of JSTOR hits mentioning O.R. was 14 percent (3,672 out of 25,708), which is akin to Africa's share of the world population (13.5 percent) and well above its share of global GDP (3 percent). Topics related to the environment, agriculture, education, transportation, manufacturing and energy were particularly common, but across a range of topics, Africa's share of O.R. publications returned by Google, Google Scholar and JSTOR compared very favorably to other regions, e.g., Latin America and Asia.

Figure 2: Searching JSTOR for "OR/MS" (or varients), public policy topics and "Africa" generated 3,672 articles at the intersection.

Despite this large body of publications, Papoulias (1984) notes that in the past O.R. models created for the developing world were often not implemented. In a study of Nigerian private companies, Ehie and Smith (1994) found that the top four "problems encountered in the use of O.R. techniques" were: (1) "insufficient trained personnel," (2) "need for software development," (3) "lack of appropriate software packages," and (4) "need for quick solutions." Inasmuch as those are the principal barriers, an emphasis on end-user modeling might be very useful for translating the energy behind the new societies and conferences into successful implementation and concrete progress on the ground.

Potential of End-User Modeling in Africa

End-User Modeling in the United States As has been discussed previously in OR/MS Today, "end-user modeling" has revolutionized the way the business world uses O.R. and quantitative modeling more generally (Plane, 1994; Grossman, 1997; Powell, 1997). Now managers need not be mere consumers of others' analysis. They can be the analyst for many non-trivially sized problems on their own laptop. Managers "live in spreadsheets"; Excel alone has over one hundred million users. Today's Excel is powerful with its built-in Solver optimization package (Fylstra et al., 1998) and abundance of inexpensive, easy-to-use add-ins for simulation, decision trees, advanced optimization and other applications (e.g., Sam Savage's Insight.xla).

End user modeling directly addresses all of the principal barriers Ehie and Smith (1994) list as reasons O.R. has not yet made more of a difference in Africa. But is end-user modeling possible in Africa? One concern might be whether Africa has the computational resources to support end-user modeling. After all, the average per capita income in Africa is about one-fortieth that in the United States. However, if computational power per dollar doubles every 18 months, then that gap translates into only an eight-year lag. Africa's ability to buy computational power today would be about what it was in the United States in 1997, and by 1997 end-user modeling had already swept U.S. practitioners and classrooms.

Another worry concerns the mathematical prerequisites. In the United States it is professional master's students who typically see a modern, spreadsheet-based modeling course. However, one does not need graduate-level mathematics to master spreadsheet modeling. Typically no math beyond algebra and exponential notation is required.

The biggest challenge might be developing the infrastructure to teach "enough" end-user modelers. How many would be enough to make a continent-wide difference? We do not know, but imagine there were one end-user modeler for every 1,000 people in Africa. (By comparison, the Bureau of Labor Statistics reports that the United States has one O.R. analyst for every 4,700 people.) One would think that size "dose" could make a difference, but Africa is the world's second most populous continent, so it would require training 800,000 end-user modelers. To show that dreaming big is not unreasonable, below we roughly estimate the investment needed to train 800,000 end-user modelers via an in-person full semester course. The bottom line annual cost of about $6 million per year is within the range of investments in human capital formation being made in Africa.

The vision would be to catalyze in Africa a revolution in management science instruction parallel to the one in North America that has been well-documented by others in these pages. The impetus for change in the United States was the 1991 revision in AACSB standards concerning the "common body of knowledge" that MBA programs had to offer to be accredited. The revision dropped any specific operations research/management science requirement (Jordan et al., 1997). That "deregulation" eliminated O.R.'s protected monopoly. The traditional O.R. course could not withstand competitive market pressures because managers are not well-served by baby versions of the algorithmically oriented classes that are the staple for O.R. majors. O.R. faculty responded with aggressive innovation. Courses were redesigned, and many new textbooks were written.

The benefits of end-user modeling education are not limited to business schools. The required core management science course at Carnegie Mellon's Heinz School of Public Policy & Management has been a "modern" modeling-based course for a decade (Caulkins, 1999). It is the foundation of a six-course sequence in end-user modeling designed for public and non-profit managers and analysts. Tailoring an end-user modeling curriculum to African data, applications and cases should likewise be possible.

Costing an End-User Modeling Training Program


One strategy for training large numbers of end user modelers would be a "train the trainers" approach (Caulkins et al., 2005). There are faculty currently teaching O.R. in Africa, albeit not primarily from an end-user modeling perspective. Suppose faculty from four African universities attended an INFORMS' annual summer workshop on teaching O.R. (Powell, 2004). Those faculty could then "train the trainers" in semester-long programs for cohorts of 25 students at a time. If the four universities trained three 25-person cohorts a year with a 75 percent graduation rate, then by the end of year three, there would be 675 trained trainers ready to teach end-user modeling. Tuition at a first-rate African university such as the University of Cape Town is $3,500 per semester, so training trainers would not cost more than about $3 million.

Table 3: Budget estimate for management science instruction in Africa Initial, One Time Costs Non-IT Costs IT Costs Total Textbook and Course Materials $385,000   $385,000 Training Instruction for faculty $75,000   $75,000 Train the Trainers costs $3,150,000   $3,150,000 IT hardware for Trainers $13,500,000   $13,500,000 Total One Time Costs $3,610,000 $13,500,000 $17,110,000         Annual Costs       Course Packets $227,813   $227,813 Salary for Trainers $2,700,000 $2,700,000 Training trainer replacements $56,250 $56,250 Hardware maintenance/ replacement $3,375,000 $3,375,000 Administrative costs $100,000 $100,000 Total Annual Costs $2,856,250 $3,375,000 $6,231,250

A salary of $4,000 per year might attract and retain excellent trainers in all but a few of the wealthiest African nations, suggesting on-going labor costs on the order of 675 X $4,000 = $2.7 million per year. If these trainers could each train 145 people per year (three semesters per year X two classes per term of 30 students each with an 80 percent graduation rate), then each year approximately 100,000 people would be trained in end-user modeling. Within a decade this training program would reach the desired one-per-thousand level of penetration.

We estimate that this training program would require an initial investment of about $17 million, including faculty training, training of trainers, course materials development and IT hardware. The annual cost is estimated at about $6 million, which includes trainer salaries, hardware maintenance and replacement, and administrative costs. So a decade-long program would cost about $80 million dollars, and the average cost of training a single modeler would be approximately $90.

IT hardware is a major cost driver assuming each trainer would need 10 laptops which would each last approximately four years. The budget presumes a standard retail price per (ruggedized) laptop of $2,000. However, purchasing at wholesale prices would cut costs significantly. If a hardware vendor donated the computers that would bring program costs down to just under $3.6 million dollars initially, and roughly $2.9 million dollars per year thereafter.

Even including full IT costs the projected cost is not large compared to the total aid given to Africa by organization such as the Africa Capacity Building Foundation (ACBF), which is financed through the African Development Bank, The World Bank and the United Nations Development Programme. For example, the ACBF provides $5 million to support a Collaborative Master's Program in Economics and recently approved $37.5 million for capacity-building initiatives in Sub-Saharan Africa.

Participants at ORPA-1 were excited about the end-user modeling approach to O.R. instruction. Their enthusiasm has been mirrored on this side of the Atlantic. Indeed, some leading textbook authors were interested in volunteering their expertise for Africa-specific text development, and software vendors were willing to donate software.

Involvement of O.R. professionals from the developed world could do more than bring down program cost. It might also be good for the profession. It is hard to imagine a better way to raise the profile of O.R. in North America than to have it become known as a driving force for progress and development in Africa. If O.R. is seen as successfully tackling some of the imposing challenges facing Africa, that would be a powerful demonstration of the promise that O.R. holds throughout the global community.

Conclusion


O.R. has a great future in Africa. There are no shortage of problems that would benefit from O.R. analysis. There are emerging communities of O.R. scholars in Africa and opportunities for dissemination to practitioners that were not viable a decade ago.

There are various ways INFORMS members can get involved, including attending ORPA-2 or IFORS-2008, sponsoring scholarships for African O.R. students, collaborating with domain experts on O.R. projects, contributing to the creation of intellectual property for an end-user modeling text or course dedicated to Africa, and/or helping to train trainers in end user modeling in O.R. Of course the specific opportunities vary by country and topic since Africa is a very large and diverse continent, but it is exciting to see O.R. beginning to spread in new ways and places.



Jonathan Caulkins (caulkins@andrew.cmu.edu) is a professor of operations research and public policy at the H. John Heinz III School of Public Policy & Management, Carnegie Mellon University, where Emily Eelman, Minoli Ratnatunga and David Schaarsmith are grad students.

References

1. Caulkins, Jonathan P., 1999, "The Revolution in Management Science Instruction: Implications for Teaching Public Affairs Students," The Journal of Public Affairs Education, Vol. 5, No. 2, pp.107-117.
2. Caulkins, Jonathan P., Emily M. Eelman, Minoli S. Ratnatunga and David K. Schaarsmith, 2005, "Operations Research & Public Policy for Africa: Harnessing the Revolution in Management Science Instruction," Carnegie Mellon Heinz School Working Paper 2005-4, Pittsburgh, Pa.
3. De Angelis, Vanda, Mariagrazia Mecoli, Chris Nikoi and Giovanni Storchi, (forthcoming), "Multiperiod Integrated Routing and Scheduling of World Food Programme Cargo Planes in Angola," Computers and Operations Research.
4. Drucker, Peter F., 1992, "Managing the Non-Profit Organization: Principles and Practices," New York: Harper Books.
5. Ehie, Ike C.; Jr., David K. Smith, 1994, "OR Utilization in Nigeria: A Sample Survey," Journal of the Operational Research Society, January, pp. 31-37.
6. Fylstra, Daniel, Leon Lasdon, John Watson and Allan Waren, 1998, "Design and Use of the Microsoft Excel Solver," Interfaces, Vol. 28, No. 5, pp.29-55.
7. Grossman, Thomas A., 1997, "End-User Modeling," OR/MS Today, Vol. 24, No. 5, p. 10.
8. Jordan, E., L. Lasdon, M. Lenard, J. Moore, S. Powell and T. Willemain, 1997, "OR/MS and MBAs," OR/MS Today, Vol. 24, No. 1, pp. 36-41.
9. Lampkin, Linda M. and Thomas H. Pollak, 2002, "The New Non-Profit Almanac & Desk Reference," Urban Institute, Washington, D.C.
10. Papoulias, Demetrios B., 1984, "Operational Research and Socioeconomic Development," The Journal of Operational Research, Vol. 35, No. 7, pp. 579-586.
11. Plane, D.R., 1994, "Spreadsheet Power," OR/MS Today, Vol. 21, No, 6, pp.32-38.
12. Powell, Stephen G., 1997, "End-User Modeling," OR/MS Today, Vol. 24, No. 4.
13. Powell, Stephen G., 2004, "Active Learning," OR/MS Today, Vol. 31, No. 4, pp.28-30.
14. Rand, G.K., 2004, "Developing O.R. for Developing Countries: the role of IFORS," Direct Connection to Developing Countries, Vol. 12, Issue 2, pp. 3-5.

Table of Contents

OR/MS Today Home Page


OR/MS Today copyright © 2005 by the Institute for Operations Research and the Management Sciences. All rights reserved.


Lionheart Publishing, Inc.
506 Roswell Rd., Suite 220, Marietta, GA 30060 USA
Phone: 770-431-0867 | Fax: 770-432-6969
E-mail: lpi@lionhrtpub.com
URL: http://www.lionhrtpub.com


Web Site © Copyright 2005 by Lionheart Publishing, Inc. All rights reserved.