Ranking transport projects by their socioeconomic value or financial interest rate of return ?

1
Ranking transport projects by their socioeconomic
value or financial interest rate of return ?

• Alain Bonnafous
• Laboratoire dEconomie des Transports
• Pablo Jensen
• Laboratoire de Physique de la Matière Condensée
  et des Nanostructures

2
How to optimize public subsidies ?
17 highway projects what is the best ranking to
maximize the total socio-economical Net Present
Value ?
Project IRRse IRR Cost (ME) Arles -
Salon 74 7,5 283 Isle Adam -
Amiens 45 6,1 488 Toulouse
Pamiers 30 4,1 450 Tours -
Alençon 28 3,6 698 Dôle -
Bourg 20 7,1 737 Saintes -
Rochefort 20 2,3 290 Tours -
Vierzon 19 4,1 800 Annemasse-Thonon
18 7,9 378 Grenoble - Sisteron 17 2,5
1880 Sens - Courtenay 15 10,1 197 Orléans
- Courtenay 13 3,0 470 Dijon -
Dôle 12 9,5 213 Lyon - Balbigny 11 1,2
770 Ambérieu - Bourgoin 11 1,2 500 Rouen -
Alençon 9 1,4 580 A88 Caen -
Argentan 9 1,2 250 Troyes -
Auxerre 4 0,9 1350
3

• Scope Maximize the total NVPse, taking into
  account a budget
• constraint F / year
• Budget is used to subsidize projects, to raise
  their IRR to 8

4

• How to rank projects ? IRRse ? IRR ? Other ?
• Does it depend on annual budget ?
• Value of F determines rhythm of completion of
  projects
• (each project draws c ? of public subsidies per
  year during
• the first 5 years).

Example for F 150, taking IRRse ranking 1st
year Arles - Salon (sub 38) 2nd year Isle
Adam - Amiens (sub 230) 4th year Toulouse
Pamiers (sub 400)

• We calculate the total NVPse returned after 50
  years.
• ? Quantify the socioeconomic value of a
  given ranking

5
IRRse always better than random !
6
IRR better than IRRse, for every F !
lt sub gt
7
Finding the optimal order

• Systematic exploration too many possibilities
  !
• Clever search
• Simulated annealing (1983)
• Genetics (1996)
• Ants (1999)

8
Example with simulated annealing atomic
cluster structure
9
Our algorithm ( simulated annealing )
Start from an arbitrary initial state
(random, IRR order...) (1) Permute two randomly
chosen projects, i and j (2) Calculate the total
NVPse for this new ranking. (3) Accept the
permutation if the new total NVPse gt old NVPse.
(4) Otherwise, calculate pexp((NVPse_new
NVPse_old)/T) (5) With probability p authorize
the permutation, otherwise cancel it. (6) If the
permutation is accepted, the temperature is
slightly decreased. (7) Start over, step (1).
10
Algorithm convergence
11
Improvement / IRRse
lt sub gt
12
Example F50 (ltFgt/10)
NVPse returned optimal 10973 O 10943 IRR
10312 IRRse 9097
Project Output Dijon - Dôle
infinite Arles - Salon 12.53
Sens - Courtenay infinite Annemasse -
Thonon 8.85 Isle Adam - Amiens 1.10
Dôle - Bourg 0.76 Toulouse Pamiers
0.69 Tours - Alençon 0.31 Tours - Vierzon
0.26 Saintes - Rochefort 0.18 Orléans -
Courtenay 0.16 Ambérieu - Bourgoin 0.08 A88
Caen - Argentan 0.05 Grenoble - Sisteron 0.18
Lyon - Balbigny 0.08 Troyes - Auxerre 0.05
Rouen - Alençon 0.06
Sub 0 35 0 9 229 163 462 833 821 520 672
1478 739 3150 2276 4986 1524
Best  analytical  ranking criterion Output O
NVPse / subsidy IRRse/(0.08 - IRR)
13
 Moral  do not spend too much on
(financially) bad projects, even if  good 
(socioeconomically speaking), otherwise you
cannot build much...

• Perspectives
• Repeat analysis on present highway programme (30
  projects)
• Full power of the method How to optimize a
  programme with interdependent projects ?
• Rapport Plan 2003  La construction dA48
  (Ambérieu-Bourgoin) interviendra dans un contexte
  de forte évolution du réseau routier national et
  autoroutier. Des aménagements programmés ou à
  létude seront soit complémentaires à lA48, en
  assurant une continuité du niveau de service pour
  certaines relations, soit défavorables à lA48 en
  favorisant les itinéraires concurrents...
  scénarii ou -