From the Channel Tunnel to Boston's Big Dig, large cost over-runs and schedule slips have characterized major transportation infrastructure projects (aka mega-projects). The best research on the extent of this problem is still the 2003 book Megaprojects and Risk, by Danish researcher Bent Flyvbjerg and colleagues. Using a large international database of 258 projects, they found that 90% suffered cost over-runs. They documented systematic "optimism bias" in the forecasts of costs, traffic, and completion date for most projects. As a remedy, they suggested that far more risk should be allocated to parties that have an incentive to reduce such outcomes, and that decisions to proceed with mega-projects should be based on "the willingness of private financiers to participate in the project without a [government] guarantee."
Flyvbjerg's database did not distinguish between projects developed under long-term concessions and those developed under traditional procurement models. But now an extremely well-done study has addressed that question, using a database of 54 infrastructure projects in Australia, of which 21 were concession-type projects (which the report defined as PPPs) and 33 were Traditional. It was conducted by The Allen Consulting Group and the University of Melbourne for the nonprofit group, Infrastructure Partnerships Australia. Link (pdf)
The findings of "Performance of PPPs and Traditional Procurement in Australia" are dramatic. "Our overall conclusion is that PPPs provide superior performance in both the cost and time dimensions, and that the PPP advantage increases (in absolute terms) with the size and complexity of projects." More specifically, on $4.9 billion of PPP projects, the net cost over-run was just $58 million, which (at 1.2%) is not statistically different from zero. By contrast, on $4.5 billion of Traditional procurements, the net over-run was a whopping $673 million (15%). Likewise, regarding on-time completion, PPP projects were on average completed 3.4% ahead of schedule, versus Traditional projects averaging 23.5% late. And while smaller Traditional projects tended to be completed on schedule, on-time performance decreased sharply as project size increased-but size made no difference in the on-time performance of PPP projects.
Two other findings are worthy of note. First, "In contrast to commonly held perceptions about the relative transparency of PPPs, we found that PPP projects were far more transparent than Traditional projects, as measured by the availability of public data." Which illustrates one of my pet peeves about critiques of tolling and PPPs-Compared to what? Critics often compare these new approaches to some kind of ideal world, instead of the funding mechanisms and institutions that actually exist.
The report also debunks "the myth of government 'risk-free' borrowing." It points out that "taxpayers ultimately and always bear the costs of cost over-runs and other project risks that cannot be assumed away" by the fact that government can borrow at a low interest rate. That is true if the alternative to PPP toll revenue-based financing is government general obligation bonds, as in Australia. In the United States, the alternative for most large highway projects is a government toll agency issuing tax-exempt toll revenue bonds. The interest rate on those bonds does bear a relationship to the project's or toll agency's financial soundness. But there is still the question of who ultimately bears the risk of optimism bias. For toll agency projects, that risk-bearer is not the taxpayer but the toll-paying customers. The annual caps on toll rates that are built into long-term concession agreements (protecting their customers) do not apply to roads built by state toll agencies. In fact, their bond covenants nearly always require them to increase toll rates as necessary to pay the agreed-upon debt service. So a version of the Australian point about risk transfer applies here, albeit in a different form.