25(34)
Stricter requirements in this respect justify the following threshold values:
Requiring that procurement pays in at least 50 percent of cases
implies that the threshold value should be about 20 percent higher.
This is equivalent to basing the design on the median rather than
on the average.
Requiring that procurement pays in at least two-thirds of cases
implies that the threshold value should be 100 percent higher.
Requiring that procurement pays in at least three-quarters of cases
implies that the threshold value should be 185 percent higher.
A 95 percent confidence interval for the average gain is obtained as the
mean ± 16 percent. It is obvious that the choice of safety margin —
basically a political choice — is far more important to the result than the
uncertainty of the estimate.
If the cost among suppliers for producing tenders is included, the optimal
threshold value is augmented by about 20 percent. Even though this could
be defended as a socioeconomically based alternative, producers seem to
consider the costs of producing tenders as part of normal business, as
remarked above. It is worth noting that the Confederation of Swedish
Enterprises in the general debate has argued for a low threshold value in
order to secure a level playing field in as large a number of procurement
operations as possible.
If the analysis had been based on the (unrealistic) strategic model used in
mainstream studies of public procurement, the optimal threshold value
would have been lower.
An Example
The trade-off between administrative costs and gains is next illustrated in
some detail for one of the procurement operations in the material, namely
the procurement of consultancy services by the Swedish Agency for Public
Management illustrated in Figure 1. The number of tenders was 27, which
is unusually high. The evaluation in this case was rather ambitious; reports
from the reference projects for each of the tenderers were studied and
evaluated by the staff of the procuring agency. The marginal cost of
examining a new tender was 0.207 percent of the estimated value of the
contract. The marginal gain from a new tender can be computed from the
order statistics by using Bury’s asymptotic formula. It turns out that the
marginal gain equals the marginal cost when the number of tenders is 27.
In other words, if the procuring agency had been able to choose the
number of tenders, it would have chosen exactly the same number as was
the result of the open tender procedure.
