Determinants of adaptation costs in procurement: an empirical estimation on Italian public works contracts

Abstract

It is widely agreed that rules governing public procurement should be designed to achieve value for money. However, in the public works sector, ‘… the good being procured is usually complex and hard to be exactly specified ex ante, … [and] alterations to the original project might be needed after the contract is awarded. This may result in considerable discrepancies between the lowest winning bid and the actual costs that are incurred by the buyer’ (Bajari et al., 2006). There is now a wide body of literature focusing on cost escalation during the execution of contracts and their estimates reveal that it can be often quite large. This article is aimed at offering an empirical test of the determinants of adaptation costs in the public works procurement. Using a detailed data set on Italian public works contracts, we run an empirical analysis, grounded on the main conclusions reached in the literature, to test for the main drivers of adaptation costs.

Keywords:

• procurement
• public work
• cost overruns

JEL Classification:

• H57
• D23
• L14
• C34

Acknowledgements

The authors wish to thank the Autorità di Vigilanza sui Contratti Pubblici di Lavori, Servizi Forniture for supplying the data used in the analysis. The authors would like to thank three anonymous referees for their helpful comments and suggestions. The usual disclaimer applies. Even if this article is the result of the joint work of the authors, Section IV is attributable to Calogero Guccio.

Notes

¹ Italy being a member State of the EU, its legislation on procurement needs to be designed according to the principles set up in the EU Directive.

² As Kelman (2002) points out, these features strongly characterized the USA procurement before the legislative reforms passed in 1994, 1995 and 1998.

³ Open procedures are the rule and negotiated procedures can be adopted only in well defined circumstances.

⁴ This is in line with the more traditional results of the economic literature on auctions. For an overview of this literature see Klemperer (1999).

⁵ The law specifies under what specific circumstances renegotiation is allowed, as well as the maximum amount, which is permitted, and the authorization procedure required. When a bigger renegotiation is needed, the contract is revoked and a new tender is issued. The project designer is liable for the mistakes in the project, which cause major revisions and, therefore, the revocation of the contract.

⁶ Expected total costs do not necessarily decrease when the number of bidders increases: the administrative costs as well as the costs for the preparation of the bid should not be undervalued.

⁷ The advantages of such a kind of relationship have been stressed by Williamson (1985).

⁸ Rizzo (2008) points out that the creation of an independent Authority to supervise and regulate the procurement market is not widespread in other EU countries and offers a tentative political economy explanation.

⁹ Autorità di Vigilanza sui Contratti Pubblici di Lavori, Servizi e Forniture (2007). The demand is not evenly distributed across the different geographical areas and the different sectors.

¹⁰ In 2006 there were 12 231 public works below 150 000 euros (38.1% of the total) and, at the other extreme, 150 works above 15 million euros. 83% of works was below 1 million euros (Autorità di Vigilanza sui Contratti Pubblici di Lavori, Servizi e Forniture, 2007).

¹¹ The effects of infrastructures on economic growth are extensively investigated in the literature (Milbourne et al., 2003; Wilhelmsson and Wigren, 2009).

¹² Flyvbjerg et al. (2002) provide a similar definition for cost overruns, which is the other term usually encountered in this literature: ‘actual costs minus estimated costs in percent of estimated costs’, where estimated costs are those forecasted at the time of the decision to realize the project. The authors are well aware, however, of the fact that estimated costs can be computed at different times: project planning, decision to build, tendering, contracting, etc. It must also be noted that the term adaptation costs is sometimes used with a substantially different meaning. Bajari et al, (2006) define adaptation costs as ‘any costs that are incurred above and beyond the direct production costs of the project’, whereas they consider the direct production costs as the other source of the cost increases following a modification of the initial plan, that is the direct costs of the additional work.

¹³ Autorità di Vigilanza sui Contratti Pubblici di Lavori, Servizi e Forniture (2006).

¹⁴ At the same time, in Italian public works delays are very striking: about two thirds of public works completed by 2005 involved delays higher than 20% of the completion time agreed upon in the contract. The determinants of such a phenomenon have been investigated by Guccio et al. (2009b).

¹⁵ Ganuza (2007) provides a rational explanation for what could be regarded as underinvestment in project design. A higher investment on a more accurate initial design lowers the probability of renegotiation and of awarding the project to the most efficient firm, but it increases its rents, when competition is not perfect.

¹⁶ Odeck (2004), however, shows that, in the case of the Norwegian road sector, cost overruns appear to be more severe for smaller projects as compared to larger ones.

¹⁷ See also Kahneman and Tversky (1979) and Flyvbjerg (2005), among the other studies that develop the idea of optimism bias.

¹⁸ Lovallo and Kahneman (2003) refer to the tendency of individuals to exaggerate their own talents, or the degree of control over events, to anchoring, that is a project plan comes out as an adaptation of an original idea, etc.

¹⁹ Flyvbjerg (2005) regards political incentives to overestimation of benefits and underestimation of costs as a separate explanation of cost overruns. However, Lovallo and Kahneman (2003) also consider organizational pressure as a source of optimism bias: ‘Every company has only a limited amount of money and time to devote to new projects. Competition for this time and money is intense, as individuals and unit jockey to present their own proposals as being the most attractive for investment. Because forecasts are critical weapons in these battles, individuals and units have big incentives to accentuate the positive in laying out prospective outcomes’. In general, therefore, optimism bias can also have a rational foundation.

²⁰ Within the context of concession contracts, Guasch (2004) has examined the potential incentive of governments in renegotiating contracts.

²¹ Ware et al. (2007) present this behaviour, also known as lowballing, as part of a colluding agreement with procurers, who would guarantee the winner a convenient renegotiation, sharing the gains, which will arise.

²² Public works are articulated in sub-categories, i.e. the different components of the overall work, which contribute, according to their relative relevance for the specific work, to the estimated total value. It is plausible to assume that the more complex is a public work the higher is the number of sub-categories involved in its implementation.

²³ Different models of governance refer to appointment methods, soft or hard budget constraints, etc. It must be noted that a different governance model may also impinge on incentives to monitoring the implementation of the work.

²⁴ A somewhat similar differentiation of contracting authorities can be found in Bandiera et al. (2009), a work analysing the behaviour of Italian procurers in the purchase of standardized goods.

²⁵ The relations of causality between the difference of cost and the procurement procedure for each contracting authority need to be explored. A lower difference of cost for a category of contracting authority might not be determined necessarily by a more careful behaviour but by a larger use of negotiated procedures (which reduces the risk of contractual revisions).

²⁶ The sample was also selected on the basis of the completeness of the records included in the data base. To limit the heterogeneity, the public works above 5 million euros were not included in the sample because of the longer time lag required to complete complex works. Moreover, public works with a final cost lower than the contract cost were not taken into account because of the lack of adequate information.

²⁷ We are unable to identifiy the panel structure at contracting authority level and assume this panel structure of the data due to the relevant difference between public work categories for adaptation cost.

²⁸ The range of the variable in the sample is between 0.0000 and 1.8037.

²⁹ To control for sample selection we also considered a Tobit II estimate (Amemiya, 1984) achieving comparable results. Estimation can be obtained by the authors upon request.

³⁰ A standard Tobit fixed effect model would be inconsistent, since the fixed effects cannot be treated as incidental parameters in nonlinear models without biasing the other model coefficients (Wooldridge, 2002).

³¹ For panel Tobit with random effects, we use bootstrap with 50 interactions (Cameron and Trivedi, 2005).

³² We employ this model because of the bias in standard Tobit estimates if an unobservable sector effect is included. We also study the pairwise correlations among independent variables and the results, reported in Table A1 in the Appendix, suggest that there are no severe problems of multicollinearity.

³³ A system for evaluating the bids, based on the notion of standard costs, might be more effective in preventing underbidding because it might reduce the asymmetrical information of the contracting authority.

³⁴  Table A2 in the Appendix.

³⁵ The only exception is the coefficient of the variable INT , which has a negative and significant effect, different from the general model, probably in relation with the very modest value of the coefficient.

³⁶  Table A3 in the Appendix.

³⁷  Table A4 in the Appendix.

³⁸ See footnote 14 in Section II.