Last week Fairfax Media reported that the Parramatta Light Rail is set to cost the NSW government not A$1 billion (together with part of the proceeds of a levy on new residential developments), as budgeted, but a whopping A$3.5 billion.*
People could be forgiven for thinking that the revelation of a A$3.5 billion price tag is surprising, and much higher than expected from what’s been budgeted. But as a new Grattan Institute report shows, cost overruns in public transport infrastructure are far too common.
The report is the first comprehensive Australian study of all 836 transport infrastructure projects valued at A$20 million or more and planned or built since 2001. It reveals that over the past 15 years, Australian governments have spent A$28 billion more on transport infrastructure than they told taxpayers they would spend. Much of this overspending should never have happened.
The A$28 billion over the originally promised costs does not stem from the accumulation of small cost overruns on most projects. Rather, most projects come in reasonably close to their original announced cost, as the figure below shows. The problem is that when projects do run over, it can be spectacular: 90% of the cost overrun problem is explained by the 17% of projects that overran their cost promise by more than 50%.
The main culprit for this startling finding is premature announcements. Ministers and shadow ministers frequently promise to build a road, bridge or rail line for a particular cost, especially in the lead-up to an election.
History shows that projects with costs announced before a formal budget commitment are at much more risk of a large cost overrun than those announced when they are better developed. Over the past 15 years, 74% of the total value of cost overruns is explained by the 32% of projects announced early.
The poor performance of projects with early announcements is not just a warning to mistrust politicians’ promises. Premature cost announcements continue to haunt projects throughout their lives, as the figure below shows.
It is not common practice to analyse projects from as far back as when they are first announced, but it should be. Once politicians have announced a project, they and the public treat that announcement as a commitment. They are right to do so: two-thirds of these projects end up being built.
In fact, it is not common practice to analyse the cost outcomes of infrastructure projects in any systematic way. Analysts are hindered by the absence of meaningful reporting on the outcomes of these projects – by the lack of information on how well they performed against the costs and benefits, such as travel time savings, that were used to make the original investment decision.
Failing to collect and aggregate information about past projects also hampers those trying to estimate what today’s projects will cost. Without actual Australian data on past projects, cost estimators may be able to factor in the prices of equipment, material and the workforce, but they can’t factor in how to protect against unusual and unforeseeable events that can have a big impact – “unknown unknowns”.
Today’s projects suggest we are not learning from history. For every project, cost estimators forecast what they expect the project will probably cost (known as a “median” or “P50” cost estimate), and also what it could cost for the one in ten times that a project will probably go over budget – a “worst case” cost estimate (known as a “P90”).
The experience of the past 15 years shows that the difference between the median cost estimate and the “worst case” cost estimate is 21% on average for transport infrastructure projects valued at A$20 million or more.
But cost estimates for current projects do not reflect this experience. Instead, the difference between the median and “worst case” estimates is generally about a third to a half this size. Judging by recent history, this suggests that either the median cost estimate is too high or – more likely – the “worst case” cost estimate is too low.
What’s wrong with this approach is that it probably won’t be one in ten projects that governments are expecting, but actually more likely one in three, that go over the “worst case” estimate.
Our research also shows that, if history is any guide, these overruns will be bigger than expected. Some of today’s big-budget projects that have these surprisingly low “worst case” cost estimates are the Inland Rail between Brisbane and Melbourne, Sydney’s WestConnex and Melbourne’s Western Distributor.
This is not to slight the motives or the abilities of cost estimators, who deal with intrinsically uncertain costs. But without historical information to draw on, these estimates are inherently limited. The evidence is clear: over the past 15 years, cost estimators expected between 10% and 25% of projects to exceed their budget. In fact, 34% did so.
Policy changes required
The solutions are within reach. Governments should not commit to building transport infrastructure before tabling a proper evaluation and the underlying business case in parliament. And once a project is completed, governments should report to the public on how it performed against the costs and benefits behind the original decision to invest.
Governments can also improve project assessment methodologies, collect and publish data that would enable cost estimators to learn from experience, and manage contingency funds with more discipline.
We must start to learn from history. Our infrastructure systems should promise what is worth having, and then deliver what is promised.
This article has been amended in the first paragraph to clarify that the original A$1 billion budget was to be supplemented with part of a new levy on new residential developments.