LONG TERM INFRASTRUCTURE RISK MANAGEMENT PROBLEMS IN THE USE OF THE NET PRESENT VALUE CRITERION

摘要

One of the problems in the management of infrastructures is the replacement of the constituting components because of ageing. Replacement is very costly, however, not replacing them but running them until failure can be very risky. In determining the optimal replacement moment two major problems exist. The first is how to compare the costs of replacement with the non financial risks. The simplest way around this is treating the non-financials as a financial risk by monetizing the effects, like Essent Netwerk does. The second problem is the method to use to calculate the optimum. For financial optimization a net present value approach is best used. However, using the NPV criterion in a replacement decision is not straightforward. One can decide to replace the asset at age x, but the asset could fail beforehand. This can be resolved by using a probabilistic approach, trying out different moments and choosing the moment with the lowest total costs. However, if an asset base consists many different asset types this is vary labour intensive, especially if a sensitivity analysis is to be performed. A simpler alternative for determining the replacement moment is the marginal approach. This means comparing the risk of failure for the next year with the cost of advancing the replacement one year. If a parameterized failure model is used, the optimal replacement moment can even be determined analytically. In theory, this should yield exactly the same outcome as the NPV approach. Surprisingly, if this moment is applied in a probabilistic model on a population of infrastructure assets, one typically finds that this does not result in the lowest NPV for the population. Analysis of this anomaly showed that this appears in case the life expectancy of the assets is long relative to the interest rate, because the present value of the risk then only builds after what is generally accepted as a reasonable cut-off point for the NPV calculation. Therefore, for long term risk management decisions it is better to use the marginal approach instead. If a Net Present Value model is used to determine the required investment programme one should be careful to take a discounting period long enough, which can be more than 100 years.