Do Cost–Benefit Analyses Influence Transport Investment Decisions? Experiences from the Swedish Transport Investment Plan 2010–21

Abstract

Cost–benefit analysis (CBA) for transport investments is particularly useful for situations where a large number of investments have to be ranked against each other. This study draws on experiences from the development of the Swedish National Transport Investment Plan 2010–21. We study how CBA results were used in the process of shaping the investment plan and what influence they had on investment decisions. In particular, we compare the planners' rankings versus the politicians' rankings. We find that planners' rankings of investments are influenced by benefit–cost ratios (BCRs), in particular for low and moderate BCRs, while the politicians' rankings are not. By interviewing planners about how CBA was used in the process, we clarify what role CBA actually played in the planning process. We find that not only did the CBAs play a role in investment selection, they also forced investment design to be more cost-efficient. Furthermore, we explore planners' implicit valuations, as revealed by their investment selection, finding that freight benefits were implicitly valued higher and traffic safety lower than the officially recommended CBA weights. Finally, we identify the most important areas for improvement of CBA state-of-practice methodology.

Acknowledgements

Pia Sundbergh, WSP, worked together with the authors in chairing the Appraisal Committee. She also provided valuable help with the quantitative analyses presented in this paper. We are grateful for helpful comments by John Nellthorp, James Odeck and anonymous referees. The research was supported by the National Road and Rail Administrations and VINNOVA.

Notes

The analysis is restricted to rail and road investments, which comprise the bulk of the plan; a few investments for sea transport are excluded, as are funds for unspecified small measures.

Formally, there are thus 1 + 21 plans, but to facilitate reading, we will refer to them as one plan with two parts. The plans also contain investments for other purposes as well as funding for maintenance, but the focus in this paper is limited to investments.

Previously, the Road Administration and the Rail Administration had carried out their planning more or less independently. This time, however, they carried out the entire planning process jointly. The Road and Rail Administrations were merged into a Transport Administration in April 2010.

The Rail and Road Administrations were already using the same transport models (SAMPERS for person traffic and SAMGODS for freight traffic) and the official national CBA guidelines for the transport sector (described in SIKA, 2008).

All investment costs in this paper are given as 2009 present values and expressed in consumer prices, to be consistent with the costs used in the CBAs (Swedish CBA is carried out in consumer prices). Hence, cost figures differ somewhat from the corresponding figures in the government budget, due to discounting effects and the conversion from producer to consumer prices.

Net benefits equal total benefits minus total costs.

This follows from the standard ‘knapsack problem’: given a set of objects with values v _i and weights w _i , maximizing the total value of a selection j of objects Σv _j , subject to a constraint on the total weight Σw _j . The solution is to rank objects according to v _i /w _i and to keep selecting objects until the constraint is met. What ranking criterion should be used in public CBA depends on what the equivalent of the ‘weight constraint’ is.

The average NBIR of all possible investments with NBIR >− 1 is 0.46, and no investments with NBIR <− 1 are included in the Initial Plan. In other words, the Initial Plan is slightly worse (in NBIR terms) than a random sample of all investments with NBIR >− 1 would have been.

The second model, however, indicates that the NBIR does have an effect on low NBIR values (below zero) in the regional part of the Initial Plan. This might be due to spurious correlation, but it may be a sign that NBIR results have been used as a ‘screening tool’ to avoid the worst investments. As we will see later on, there is evidence that CBA results are often used in this way in other situations.

The role of CBA results in this process is an interesting research field in itself, but it is not within the scope of this paper. Instead, we concentrate upon the selection from this list into the plan proposal.

In the planning directives, the government stated that at least 50 percent of the investment budget should be used for roads. The planning committee decided early on to make a proposal where the investment budget was split 50:50 between rail (including sea) and road.

This is for many reasons, but one important reason is the lack of model management. Databases of socioeconomic data, networks, etc., are not sufficiently updated and debugged, often giving rise to confusing erroneous results. This may very well be the single most important issue to improve in CBA practice.

‘Densely populated regions’ consist of the counties where the major cities are found, i.e. Stockholm, Västra Götaland and Skåne. Investments in purely rural areas of Västra Götaland and Skåne have been put in the category of ‘other regions’. ‘Sparsely populated regions’ consist of the counties Norrbotten, Västerbotten, Västernorrland, Gävleborg and Dalarna.

Note that they also stated that CBA results should affect the allocation between rail and road investments. This could either be interpreted as a request to allocate more to road if those investments were to show higher NBIRs than rail or as a memento that the Government was prepared to change the balance when finally deciding upon the National Transport Investment Plan.

Excluding the Initial Plan, i.e. the politicians' investment selection.