Improving Mid-Term, Intermediate, and Long-Range Cost Forecasting for State Transportation Agencies (2020)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

17 2 State-of-the-Practice of Cost Forecasting 2.1 Introduction This chapter summarizes the state-of-the-practice of cost forecasting in the transportation construction industry. The chapter is focused on current practices, needs, and opportunities for improvement in mid-term (3 to 5 years), intermediate-range (up to 15 years), and long-range (more than 15 years) cost forecasting. The information presented and analyzed in this chapter was collected through a comprehensive literature review, an online survey administered to state transportation agencies (STAs), and feedback provided by the AASHTO Technical Committee on Cost Estimating (TCCE), which served as an Expert Advisory Panel (EAP) for this study. Delivering the online survey to potential participants was particularly challenging since the forecasting activities considered in this study correspond to multiple transportation programs, which are usually handled by different offices/groups within STAs, making it challenging to identify staff members familiar with the cost estimating process across planning phases. Likewise, targeting various offices per agency was also challenging. A total of 20 responses were received through the online survey platform, including one from a local transportation agency (Contra Costa Transportation Authority [CCTA], California). Further efforts were made to collect relevant policy documents from all STAs in order to facilitate a better understanding of current transportation planning and cost forecasting practices. Manuals and standard procedures from all 50 STAs were reviewed in this study, including documents from agencies that also completed the survey. Figure 2.1 shows the STAs that responded to the online survey. Figure 2.1 Survey Responses and Policy Documents Reviewed

18 2.2 Transportation Planning Programs and Cost Forecasting Periods Mid-term (3 to 5 years), intermediate-range (up to 15 years), and long-range (more than 15 years) cost forecasting are generally associated with State Transportation Improvement Programs (STIPs)/Transportation Improvement Programs (TIPs), Intermediate-Range Plans (IRPs), and Long-Range Transportation Plans (LRTP), respectively. However, survey responses and policy documents showed some variability in the length of those programs. It should be noted that the length of a program is equal to its respective cost forecasting period. Some variability was also observed in the length of Transportation Asset Management Programs (TAMPs), Bridge Management Programs (BMPs), and pavement management programs (PMPs). Figure 2.2 illustrates the variability in the forecasting periods for different transportation programs and the different updating frequencies adopted by STAs across the country. Numbers at the top and bottom of each bar correspond to the maximum and minimum values (in years), respectively, provided by survey respondents. Figure 2.2 Forecasting Time Horizons and Updating Frequency of Transportation Programs Forecasting periods and program updating frequencies are usually regulated by federal, state, or local statutes, which establish minimum standards that must be met or exceeded by STAs. For example, federal regulations establish that LRTPs must be developed with a minimum 20-year forecast period (23 USC §135 2018), allowing time periods longer than 20 years. This study has also found 30-year LRTPs prepared by some STAs, such as the California Department of Transportation (Caltrans 2006). Likewise, the current LRTP implemented by the Connecticut Ti m e  Ho riz on Up da tin g  Fr eq ue nc y Ti m e  Ho riz on Up da tin g  Fr eq ue nc y Ti m e  Ho riz on Up da tin g  Fr eq ue nc y Ti m e  Ho riz on Up da tin g  Fr eq ue nc y Up da tin g  Fr eq ue nc y Up da tin g  Fr eq ue nc y Up da tin g  Fr eq ue nc y Ti m e  Ho riz on Ti m e  Ho riz on Ti m e  Ho riz on LRTP IRP STIP TIP TAMP BMP PMP 0 5 10 15 20 25 30 35 20 30 10 4 5 12 5 10 4 4 10 4 4 56 20 20 20 2 25 1 0.5 1 1 1 1 1 Fr eq ue nc y

19 Department of Transportation covers a 32-year period from 2018 to 2050 (CTDOT 2017). The time period for LRTPs may also vary within an STA. For instance, the current LRTP of the Georgia Department of Transportation (GDOT) covers a 25-year period, from 2015 to 2040. It was developed to replace GDOT’s 30-year LRTP with goals and infrastructure needs to be addressed between 2005 and 2035 (GDOT 2016). Similarly, federal regulations state that STIPs must “cover a period of no less than 4 years and shall be updated every 4 years, or more frequently” (23 USC §135, 2018). Some STAs have decided to exceed these minimum federal requirements by implementing longer STIPs and more frequently updating schedules. This is the case of the STAs in New Jersey and North Carolina, whose 10-year STIPs are updated every 2 years (NJDOT 2018; NCDOT 2018). The Wisconsin Department of Transportation’s STIP is updated every six months, and its forecasting periods may vary from 6 to 10 years. STAs may decide to exceed federal requirements by implementing longer forecasting time periods and/or more frequent updating schedules in order to either meet stricter state or local regulations or because it could better fit their planning needs and management practices. For example, longer STIPs may be motivated by the need to facilitate a better transition from the LRTP to the STIP without implementing an IRP. Federal and state statutes associated with transportation programs seem to be less restrictive for TAMPs, BMPs, and PMSs regarding required time horizons and updating cycles. Forecasting periods for those programs range from 5 to 20 years. Such level of variability in program lengths and updating periods means that guidelines and cost forecasting methodologies proposed by this study should have the flexibility to handle different scopes of work along with a wide range of possible forecasting time horizons—an observation that was validated and stressed by the TCCE. 2.2.1 Configuration of Program Cost Estimates In addition to the different forecasting time horizons involved in transportation planning, STAs must also deal with great variability in the configuration and level of detail in the content transportation programs. Federal and state statutes offer general guidance on the required content and structure of transportation programs, providing STAs with some flexibility regarding the level of detail of the strategies, methods, and anticipated projects included in the programs. That level of detail is also reflected in the configuration of financial plans and cost estimates. Figure 2.3 summarizes survey responses on the configuration of cost estimates for each type of program. This figure shows how the cost estimate configuration tends to move from a lump sum approach into a more detailed itemized estimate as infrastructure needs and projects move from LRTPs into STIPs/TIPs. The online survey did not show a clear trend in the configuration of cost estimates in TAMPs, BMPs, and PMPs, which could be explained by the wide range of possible forecasting time horizons associated with those programs.

20 Figure 2.3 Cost Estimate Configuration per Transportation Program – Survey Responses (Survey Responses = 18) 2.3 Forecasting Tools and Methods Survey participants were also asked to indicate what forecasting tools and methods are used by their agencies for cost forecasting purposes in mid-term, intermediate, and long-range planning. Responses to this question are summarized in Figure 2.4. This figure shows that the outsourcing of cost forecasting services does not seem to be a common practice among STAs. On the contrary, historical bid data is more commonly used to define market trends, which is also the approach adopted by this study. Figure 2.4 Cost Forecasting Methods and Tools (Survey Responses = 17) 0 2 4 6 8 10 12 LRTP IRP STIP TIP Fre qu en cy Lump Sum Cost Estimate Itemized Cost Estimate by Type of Work Itemized Cost Estimate by Type of Cost Itemized Cost Estimate by Location Itemized Cost Estimate by Other 2 15 8 5 2 6 8 4 0 9 4 3 4 0 5 00 7 4 3 4 0 4 00 2 4 6 8 10 12 14 16 Ou tso urc ed Co st Es tim ati ng Se rvi ces His tor ica l B id Da ta Ma jor Co st I tem s u sin g Sta nd ard ize d S ect ion s Inp ut fro m a P an el o f Ex pe rts Pa ram etr ic E sti ma tin g Co st- Ba sed Es tim ati ng Ris k-B ase d E sti ma tin g Re gre ssi on An aly sis FR EQ UE NC Y ( SU RV EY RE SP ON SE S) Mid-Term Intermediate-Range Long-Range

21 Using a panel of experts to provide input during the forecasting process is a formal way to bring valuable staff experience and knowledge into this process. Usually, these panels are strategically formed on a program-by-program basis or for major capital projects, attempting to cover all knowledge areas and disciplines required by each program/project. Given that the survey showed low use of outsourcing services associated with cost forecasting, it is reasonable to assume that panels of experts supporting cost forecasting efforts are mostly formed by in-house experts. After historical bid data, risk-based estimating seems to be the second method most commonly used by STAs in cost forecasting over long time horizons among those methods shown in Figure 2.4. “Risk-based estimates produce an expected value and a range of project costs. […] Estimators will typically use risk-based estimates during the planning, scoping, and early design phases” (Molenaar et al. 2013). The capacity to account for estimating risk in the cost forecasting process was added by this study in the form of the risk-based forecasting timeline discussed in Section 1.3. Even though regression analysis techniques could be easily paired with historical bid data (the most common source of data) to model market trends and forecast those trends into the future, those do not seem to be a popular approach among STAs. With good quality historical bid data, an STA could develop regression models to represent cost trends in the form of equations that facilitate the estimation of construction costs for different forecasting time horizons. One of the reasons that could explain why STAs stay away from regression analysis could be the fact that it may require a certain level of knowledge in data analytics, economics, or statistics –skills not commonly found in STAs, as showed in the survey responses illustrated in Figure 2.5. Figure 2.5 STA Staff with Economics and/or Statistics Background (Survey Responses = 17) Chapter 5 and NCHRP Research Report 953 include straightforward guidelines on regression analysis. Likewise, the Cost Forecasting Toolkit includes a spreadsheet to facilitate regression analyses on CCIs without the need of skilled statisticians or data analysts. It should also be noted that this report and the guidebook include instructions for the use of a novel cost forecasting methodology proposed by the research team called Moving Forecasting Error (MFE). This

22 methodology proved to be effective at producing forecasted values from the analysis of historical bid data and is the one that produces outputs in the form of risk-based forecasting timelines. The survey also asked STAs about information technology (IT) resources used to aid cost forecasting during early transportation planning stages. Figure 2.6 shows the responses received for this question, making it clear that Microsoft Excel is the tool most commonly used by STAs for cost forecasting purposes, followed by in-house estimating software. In order to take advantage of the familiarity of STAs with Microsoft Excel, the research team decided to use that software as the interface for the Cost Forecasting Toolkit associated with this report. Figure 2.6 Information Technology Tools used in Cost Forecasting (Survey Responses = 17) AASHTOWare is a comprehensive software package divided into multiple modules intended to assist transportation agencies with planning, design, construction, and contract administration activities. Although AASHTOWare has long-range cost forecasting capabilities (AASHTO 2017), it seems to be more commonly used for mid-term forecasting. However, Figure 2.6 suggests that STAs still prefer the use of Microsoft Excel spreadsheets or their own estimating software to forecast construction costs at all planning levels. Figure 2.6 also shows that statistical software packages are rarely used by STAs, which could be explained by the lack of staff with the required skills to use this type of software, as shown in Figure 2.5. 2.4 Current Use of Construction Cost Indexes and Inflation Rates The study found that external CCIs are widely used by STAs for contract price adjustments or to support other cost estimating tasks, but they are less commonly used in cost forecasting as a means to develop inflation rates for mid-term, intermediate, or long-range forecasting. Among those agencies using CCIs for cost forecasting purposes, there are some using the Consumer Price Index (CPI) and the Personal Consumer Expenditures (PCE) price index. Those are the two macroeconomic indexes mentioned before in Section 1.4. As already mentioned in that section,

23 the CPI and PCE price indexes are not intended to be used at the microeconomic level, such as to forecasting transportation construction prices at the program o project level. They do not use construction-related inputs, making them unsuitable to represent the construction market. Issues associated with the use of macroeconomic indexes in cost forecasting have been recognized by some STAs, such as the Michigan Department of Transportation (MDOT), which argues that “construction inflation generally outpaces consumer inflation” (MDOT 2016). This implies that the use of the CPI or PCE could lead to an underestimation of future costs. MDOT has formed a team to investigate effective methods to develop inflation factors based on construction inputs instead of consumer goods and services. In the meantime, MDOT has decided to use a 4% percent inflation rate for cost forecasting, as recommended by the Federal Highway Administration (FHWA), when better information or methods are not available (FHWA 2017). Most STAs seem to be using standard inflation rates like this one, following recommendations from the FHWA, other federal or state agencies, or financial consultants. Those standard rates are most likely calculated by external parties from the analysis of CCIs, but they are usually not revised on a regular basis and are not directly associated with the local construction maker of each STA nor to the scopes of work under consideration. The FHWA also discusses the possibility of using external CCIs to estimate inflation rates, but also in the absence of better information or methods. In this study, an external CCI refers to a cost index not developed by or exclusively for the agency. Some examples of external CCIs are those published by the Bureau of Labor Statistics (BLS), the Bureau of Economic Analysis (BEA), the Engineering News-Record (ENR), and the FHWA itself, which publishes the quarterly National Highway Construction Cost Index (NHCCI). There are also agencies, like the STAs in Washington State (WSDOT) and Wisconsin (WisDOT), which are developing inflation rates based on external transportation-related CCIs. This is done using the services of a financial consultant (WSDOT 2018; WisDOT 2018). Although the use of a transportation-related CCI is a more appropriate approach to assess inflation trends in the transportation industry, WSDOT and WisDOT are still using an external national index to forecast local construction costs. This may not accurately represent the local transportation construction market. “Pricing changes in any single state can be affected by influences that are muted or lost in national prices and price indexes. One example of this is contractor competition, which has a strong influence on prices but has only a local or regional effect” (Molenaar et al. 2013). The previous statement points back to FHWA’s top recommendation for STAs on the importance of investing in the development of in-house CCIs with their own historical cost data for a more appropriate determination of inflation rates and more accurate cost forecasts. As occurs with the external CCIs, a number of STAs have developed their own cost indexes, but only in a few cases, they are used to support forecasting efforts over long periods of time. According to survey responses and the literature review, some STAs using in-house CCIs in their mid-term, intermediate, and long-range cost forecasting activities are Ohio, Florida, South Carolina, Caltrans, and Iowa.

24 Even though the use of in-house CCIs is the preferred approach recommended by the FHWA, the proposed cost forecasting guidelines include the use of an alternative cost indexing system call a Multilevel Construction Cost Index (MCCI). As explained in Sections 1.3 and 1.4.1, MCCIs are designed to overcome some limitations associated with the implementation of traditional CCIs, providing greater flexibility to adapt to specific scopes of work and offering greater accuracy in the tracking of construction market fluctuations. More information about the MCCIs is presented in the next chapter, Section 3.6, and in the Transportation Cost Forecasting Guidebook. 2.5 Risk Analysis Practices in Cost Forecasting Figure 2.7 summarizes survey responses on the formality of risk assessment procedures performed by STAs for mid-term, intermediate, and long-range cost forecasts. It is clear from this figure that risk management procedures tend to be less formal for long-range estimates. There is an increase in the level of formality of risk assessments as a program moves into intermediate-range and mid- term planning phases. It should be noted that some of the STAs that responded positively to the implementation for formal risk management procedures indicated that these procedures are only performed on major capital projects or selected projects included in transportation programs, and when performed, the robustness of the analysis may vary based on project size and complexity. The study found that formal risk analyses are mainly performed at the project level. Figure 2.7 Compounded Interest Rate vs. Simple Interest Rate (Survey Responses = 16) There is great variability in the robustness of risk analyses among STAs, with the STAs in Florida and Washington State on the high end of that spectrum. These two agencies count with detailed, in-depth risk management procedures, including the use of three-point estimates to generate risk- based outcomes. Those types of estimates account for estimating uncertainty in the form of triangular probability distributions defined by three parameters: minimum, most likely, and maximum estimate. The proposed Moving Forecasting Error (MFE) method described in this report and NCHRP Research Report 953 is presented as an alternative and more effective method to generate risk-based forecasting outputs. However, the Cost Forecasting Toolkit allows the use 54% 56% 44% 25%31% 33% 37% 25% 15% 11% 19% 50% 0% 10% 20% 30% 40% 50% 60% Formal risk analysis using objective standardized procedures and/or quantitative/statistical techniques Formal risk analysis using subjective knowledge based on the experience and professional judgement of a panel of experts Informal risk analysis is performed by the estimator or other individual No risk analyses are performed for these forecasting procedures Mid-Term Intermediate-Range Long-Range

25 of three-point estimates to account for estimating uncertainty in current-dollar estimates. Thus, the risk-based forecasting timeline generated by the MFE could also consider pre-forecasting uncertainty if a three-point current-dollar estimate is used. Even though the toolkit allows the use of three-point current-dollar estimates, neither this report nor NCHRP Research Report 953 provides guidance on the generation of such estimates since they are out of the scope of this study.