The World Bank is in the process of reforming its procurement system, the set of rules that borrowers have to follow when they use Bank financing to buy goods and services. Most of the proposals sound very sensible: much less “prior review” of the process for smaller contracts (World Bank staff looking over bid documents, evaluation reports, and contract documents before they are finalized); more flexibility to use other people’s procurement systems if they’re high quality; more flexibility to use quality alongside cost in evaluating bids in return for greater transparency. There’s some worrying potential expansion of an already clunky safeguards system into procurement, but that aside it all looks great — at least in theory.
It is only great in theory because we have pretty much zero evidence on what actually works and what doesn’t when it comes to the current set of World Bank procurement rules. There’s some doubt that they prevent corruption (their primary aim), but almost no empirical evidence on that topic, let alone their broader impact on development outcomes.
There’s Great Data to Analyze…
What makes this dearth of evidence particularly odd is that there is a rich set of data on procurement processes sitting on World Bank servers: information on thousands of contracts and projects that could help guide a discussion of reform. It isn’t all in the public domain (though, kudos to the World Bank, more and more of it is). Surely there’s a spare economist somewhere amongst the 10,000 plus staff in the institution who could spend a few weeks putting together and analyzing a dataset. Here’s the kind of thing it could include:
From existing research: evidence on country characteristics, project characteristics, project development outcomes, supervision, and task manager quality.
From the World Bank procurement database on prior review contracts: measures of number and size of (prior review) contracts in a project, the number of bids on contracts in the project, the average negotiation length on contracts in the project, the number of contract amendments in the project.
(More of a reach) From the INT database of investigated and tainted contracts regarding fraud and corruption: a list of contracts.
…And it Could Help Answer Some Interesting Questions
I can’t think of a natural experiment that would allow for strong causal statements out of such a database, but a bunch of correlations alone would put us in a considerably stronger empirical position than we are today. The type of questions that could be informed by such an exercise:
What makes for good procurement outcomes in terms of plentiful bids, short negotiations, and few amendments? Is it something about the country, the sector, the task manager, or the project (or all of the above)?
Do procurement outcomes matter for project outcomes? Which ones?
Do higher rates of prior review or fewer contracts in a project or (even) contract suppliers from particular regions make for better project outcomes (and/or better procurement outcomes)?
(More of a reach) Do higher rates of prior review or fewer contracts in a project, or country, sector, task manager, or contract characteristics, correlate significantly with investigated or tainted projects?
There are surely more things that could be put in the database and more questions that could be addressed (thoughts very welcome in the comments). But for an institution that prides itself on evidence-based policymaking, the Bank should surely put the horse before the cart with a little bit of analysis before implementing procurement reforms?