Public Sector Governance and Accountability Series: Budgeting and ...

348 Stephen B. Peterson
The example of the automation of the Ethiopian budget and accounts
system provides five key lessons in automating financial reforms in developing
countries:
1. Institutional factors are far more important than the technical choice in
determining the outcome of automation.
2. Information technology should not be the driver of financial reform.
Indeed, if it had been, the Ethiopian reform would probably not have
been implemented.
3. There is no a priori technical reason to favor either an OTS or a custom
solution: the choice depends on the circumstances. However, the opportunities
offered by a custom solution for learning by doing and for creation
of ownership provide strong arguments to balance the putative
advantages of an OTS solution.
4. Effective project selection and management are major factors in the
success of automation.
5. A financial and social cost-benefit analysis should be undertaken in
reviewing a policy of introducing or continuing with a customized system
or upgrading to an OTS solution.
Lesson 1: Important Institutional Factors in Determining the
Outcome of Automation
The literature contends that success of an IFMIS depends on strong high-level
commitment and support (Diamond and Khemani 2006). Such commitment
is not always to be found in developing countries, where bureaucratic rivalry,
limited technical competence at the top, and reluctance to change (which is
often well founded) may all be factors. The example of Ethiopia demonstrates
that even when “saints” (Peterson 1998b) and high-level commitment exist,
they do not ensure effective reform, because the middle-level officials have to
implement the change.
The DSA Project managing the Ethiopian reform of budgets, budget
planning, and accounts faced benign neglect, and in many ways, as described
earlier, this situation benefited the reform. Optimal obscurity can be a key
factor in the success of development projects more broadly. Projects that are
optimally obscure do not have the high expectations and scrutiny of highly
visible and political projects, which spares them criticism and unrealistic
time frames. Optimal obscurity allows projects to learn by doing, to make
mistakes, and to progress incrementally. 31 Large-scale OTS IFMISs are not
optimally obscure. They are hugely expensive, which alone makes them