Texas DOT has announced it is dropping the toll concession approach for the next stage of the Grand Parkway (TX99) north of Houston in favor of taking the traffic and revenue risk itself and using a design-build (DB) contract. In the usual brisk Texan fashion the decision came about a week after TxDOT had listed groups qualified for both concessions and DB on around 38 miles, 61km (segments F1, F2, and G of an eventual 180 mile, 290km circumferential road. It varies between about 23 miles to 40 miles, 37km-65km out from central Houston.

The 38 mile 2×2 lane segments that are part of this project run from the US290 northwest of Houston generally east through the far northern part of Harris County crossing I-45 and passing into Montgomery County to terminate at US59. There are three expressway-to-expressway interchanges with those long and expensive direct connectors and some segments with free frontage roads but the right of way will allow point upgrades where traffic justifies it.

Project cost for the F1, F2, and G stretch is put at around $1,400m – for a cost of $38m/route mile and about $9.5m/lane-mile. (TxDOT has a lower project cost number in a presentation for stakeholders: $1,129m.)

In a presentation last year (link below) TxDOT said they expected to get a concession fee to be paid at close of $1,100m. Design-build they said then would require a subsidy of $550m over toll revenue bonds.

A spokesman said that the difficult financial climate and weak traffic and revenue forecasts for the Parkway meant they were so unlikely to get strong toll concession proposals it wasn’t worth pursuing that track.

Earlier forecasting in a presentation showed 2021 revenue for F1/F2/G segments in the range $130m to $260m, 2026 $190m to $480m, 2031 $230m to $650m, 2041 $300m to $1,120m (based on read-off of a graph.)

But there’s a E segment (nearly 16 miles,25km long) already under construction that TxDOT will toll along with the three new segments.

Toll configuration

The new project segments F1, F2, G involves six mainline toll points and ten pairs of ramp toll points. The existing E segment to the southwest of F1 involves two more mainline toll points and nine ramp pairs. So E:G has 8 mainline toll points and 19 pairs of ramp tolls.

All-electronic tolling by transponder or license plate camera image is proposed, as in all other new toll facilities in Texas.

Traffic and revenue projections

Wilbur Smith’s latest traffic and revenue forecasts for the Grand Parkway are only released in the form of a spreadsheet, but they incorporate an interesting approach in which they express different forecasts with four different probabilities: P10 meaning there is only a 10% probability the forecast will not be met (90% probability of meeting them), P25 a 25% probability of failing…., P75, and P90 a 90% probability that the forecast won’t be met. And they forecast each constrained (by limited lanes) and unconstrained (with more lanes).

The forecasts use a starting toll rate of 17.1c/mile, 11c/km the toll rates increasing with the escalation of risk.

The P25 forecasts are for traffic starting in 2016 at 47.74m/year or 131k/day.

Remember this is the toller’s definition of traffic – number of toll transactions. Since point tolling is being used each trip often generate multiple toll transactions, between one and nine transactions as tolls are struck along the pike.

Traffic defined as toll transactions about doubles after the first five years (2021) to 248k/day – the ‘ramped up’ traffic.

Over the next five years to 2026 the increase is just under 30% on the ramped-up 2021 volume and 15 years later in 2041 it is up 2/3 on 2021 traffic.

Revenue rises more steeply of course because toll rates are assumed to rise, from $55m in the first full year to $119m ramped-up in 2021, then to $170 in 2026 and about $300m in 2041.

Five bid for concessions, seven for DB

Five well-known international groups based in Europe and Australia qualified to make toll concession proposals: Vinci Concessions SAS, Macquarie Ltd, Cintra Infraestructuras SA, ACS SL/Hochtieff PPP, and OHL SA. But no US group qualified, nor several overseas groups that often bid.

Seven groups have qualified for the design-build project. Some of the companies partnering with toll concession proposers were also in groups making design-build bids, but most of them were separate. (see table nearby)

TX99 was one of eleven priority projects authorized by the legislature (SB1420) in 2011 for toll concessions.

The executive director Phil Wilson is quoted in a statement recently: “We review and manage each one of these priority projects based on their own merits and think the design-build option is the best way to go forward on the Grand Parkway.”

They hope to award the contract before the end of 2012 and break ground on the project early 2013.

Probabilistic forecasts comment by T&R consultant

Robert Bain, the leading independent consultant on traffic and revenue forecasting (RBconsult Ltd www.robbain.com) says probabilistic numbers are a useful addition to forecast modeling. He says the uncertainty range as presented by WSA “does appear to be reasonable.”

He advocates another approach as a supplement.

“On the chart I plot WSA’s base case and their ‘P10’ case. P10 simply means that, by their reckoning, there’s only a 10% chance of traffic (or transactions, as it appears here) falling below these estimates.

“I also plot – in red – my own estimate of the lower boundary likely to apply to the prediction intervals associated with traffic forecasts.”

He defines two ‘states’: dynamic and stable.

“A stable environment is one in which traffic conditions are mature and the ‘drivers’ of growth (population, employment etc.) are likely to evolve in predictable ways. As traffic forecasts are likely to be more accurate under these stable conditions, the predictive interval boundary lies closer to WSA’s base case.

He also models statistically what he calls a dynamic environment – “one experiencing on-going change, where traffic forecasting is likely to be less likely. Both of my boundaries have been estimated through countless reviews of toll road traffic forecasting accuracy that I’ve undertaken over the past 10 years. And they’ve been validated through a survey I conducted of traffic forecasters (and others) asking specifically ab out the forecasting accuracy associated with time horizons of different lengths.”

Bain says WSA’s P10 sits pretty comfortably between his two (lower) boundaries – suggesting that their estimates of traffic forecasting uncertainty align with mine, and with a good deal of empirical evidence.

“Well done WSA. Traffic forecasters are generally inclined to overestimate the predictive capabilities of their models – but not he