**4. Discussion of the Statistical Analysis—The Relevance of Regulations and Norms**

We performed a simple regression to test whether the time taken to award a contract correlates with the cost of the project.

We tested this correlation on:


While the statistical significance of each variable is robust at 1% level, the adequacy of the model is unsatisfactory, with the R2 adj less than 10% (The R<sup>2</sup> adj. is 1.93% for the whole sample, 4.04% for works costing less than 1 mil € and 0.00% for works costing more than 1 mil €. Given the scarce magnitude of the R<sup>2</sup> adj. we omitted the *p*-value for the independent variable.) This means that the variables do not contribute to explaining the phenomenon. In other words, in the sample considered here, there is not only no causal relationship but also no statistical correlation between time overruns and costs of the works. These results confirm the findings of other empirical research assessing cost and time overruns as separate issues [13–15]. We obtained the same results for the whole sample and when testing the model separately on the two clusters. Given the high variability of the data, we performed the same statistical analysis, clustering the sample by the contractor. We subdivided the sample into two groups—the first one contains works awarded by municipalities (n. 4080 works) and the second one, works awarded by the other public authorities (n. 701 works). Again, the results of regressions are unsatisfactory (The R<sup>2</sup> adj. is 2.37% for the cluster with works awarded by municipalities and 1.01% for the works awarded by other public authorities. Given the scarce magnitude of the R2 adj. we omitted the *p*-value for the independent variables.).

As the dataset covers a lengthy period during which the legislation on the contract awarding procedure changed significantly three times, we divided the sample differently to check for bias relating to changes in the regulations. We considered three clusters:


The first cluster is the smallest, accounting for less than 1% of the sample. The second, covering the longest period makes up 76% of the whole sample and the third cluster includes 26% of the total dataset. The first cluster is too small to provide any robust statistical results while analyzing the other two clusters confirms the previous results. When we tried considering other independent variables—the cost of the works, the dimension of the public authority and the type of the works—the correlations are not strong enough to explain the time take to award the contracts The R2 adj. is 10.87% for the period 1999–2006, 3.02% for the period 2006–2016 and 4.45% for the period 2016–2018. Given the scarce magnitude of the R<sup>2</sup> adj. we omitted the *p*-value for the independent variables. Given the substantial diversity of the works by type and category, we refined the sample again to test for the heterogeneity of the sample. Thus, we clustered the sample considering just the two most frequent categories of works, that are "Road infrastructures" (n. 1365 works) and "Schools and buildings for social activities" (n. 896 works) and we performed a multivariate regression for each group using cost of works and the dimension of public authority as independent variables. However, again, also in this case, the results are inadequate to demonstrate a statistical correlation between the award time and cost of works The R<sup>2</sup> adj. is just 0.75% for the data of road infrastructures and is 5.04%. Again, we omitted the *p*-value for the independent variables.

This result partly confirms the previously-mentioned literature [12–15,45]—the cost and the duration of public works, and—in the present sample - the time taken to award the contract is unrelated. The dimension of the public authority is also unrelated to the time taken to award the contract, even though the contractor's specialization is recognized as a crucial feature affecting the efficiency of the contract awarding procedure.

So what decides the amount of time it takes to award a contract? Clustering our sample of data by the abnormal offer cutoff shows that current Italian regulations on the matter, *ceteris paribus*, significantly influence this variable. This is also confirmed when we compared the first group of clusters (by costs) with the second ones (by time)—what does not change between the groups and represents the different regulations governing the contract awarding procedure, is the cutoff for abnormal offers (which is €1 million).

In short, the procedure needed to deal with the formalities associated with the examination of abnormal offers may double the time it takes to award a contract (244 days against 479 days for works over €1 million costs), in the sample considered here at least, as it is shown in the descriptive statistics. In fact, the procedure is time-consuming but also paves the way to litigations between bidders, that very often lead to the contract awarding procedure being suspended and/or the winner being changed by a decision of the courts. The Italian Government [40] also recognizes that the "passage time" takes up a mean 45.7% (range: 40.5% to 55.3%) of the time needed to complete contract awarding procedures. The relevance of procedures required by law is a feature of time overrun scarcely studied. The results are underpinned by the several statistical tests performed—none of them have

demonstrated a statistical correlation between award duration and cost, nor for the whole sample, nor for the diverse cluster tested. The findings confirm the part of the existing literature that considers time overrun unrelated with cost. This feature may not be valid for each phase of the duration of the works—during the construction phase, an increase of the costs is most of the time correlated with a time increase, due to the above-mentioned design errors and underestimation of risks, that is, in the construction site. Focusing on the procedure's duration, costs (increasing or not) may be irrelevant to the time increase, as the analysis here presented confirm. The compliance of estimated time depends on the requirements to perform the award procedures. These requirements do not change according to the kind of contractors, because they vary just for the *ex-ante* estimated work cost (above and below the 1 million euros threshold, as shown in Figure 3). Some public authorities, that is, small municipalities, do not have the technical expertise and the minimum staff too, to accomplish in time the procedural steps. Also, this kind of municipalities may be involved in the development of major works and/or infrastructures that they are not able to manage in time properly. Nevertheless, further and in-depth analysis is needed to extend the finding, here limited to a specific Italian Region. To generalize the result, the findings should be validated for Northern Italy at least.

### **5. Conclusions**

Although there is plenty of literature on cost overruns in major and mega-projects, it does not seem to have improved the situation concerning the increase in costs and delays in completing public works. Public works are a relevant part of the National public economy and the efficiency of these investments contributes significantly to the sustainability of public accounts and finances. Moreover, the strategy adopted in public works may increase the spread of sustainable development and behavior at the local and territorial level. While there are several factors affecting the completion of public work—most of them in the design and construction phase, that is, at the beginning and end of the process—very little attention has been devoted to what happens in between. The present work considers the time taken to award contracts, which cannot be explained by considering the cost overrun (incurred largely during the construction phase, after the contract has been awarded). We provided empirical findings on the time taken to award contracts in a sample of 4781 public works projects planned and implemented from 1999 to 2018 in the Veneto Region, in north-east Italy, in an effort to find an explanation for time overruns in these procedures.

Our elaborations show that even small projects, defined here as public works costing less than €100 million, are liable to time overrun and this stems mostly from the administrative demands of laws and regulations on public works [46,47]. In particular, when we divided our sample into works using the €1 million cutoff for monitoring abnormal offers, the time taken to award public works below the cutoff is 244 days, while it almost doubles to 479 days for those above the cutoff (see Figure 3).

To check whether this difference in timeline is coincidental, we performed several statistical multivariate regressions where the time taken to award contracts is the dependent variable and the independent variables are other features conventionally assumed to affect time overruns, that is, the cost of the works, the dimension of the public authority and the category of the works.

**Figure 3.** The effect of the abnormal offer threshold on time taken to award contracts.

We tested this model with five different specifications—in the whole sample; in two clusters by contract value (below vs. above €1 million); in two clusters by period (2006–2016 vs. 2016–2018); in two clusters by contractor (municipalities and other public authorities) and lastly testing just the two major category of works in the sample, which are "Road infrastructure" and "Schools and buildings for social activities."

All our statistical analyses demonstrate that the longer time taken to award contracts is unrelated to the cost of the works or the dimensions of the public authorities involved (as a proxy for their contract awarding expertise). This means that the extension of the time taken to award contracts can be attributed entirely to the related procedures required by law and especially to the checks on abnormal offers.

These findings point to the need for further research. To be generalizable, our findings would need to be tested on a larger sample of data and in different regions. The same set of data could also be useful for examining other phases of the public works process, such as design and construction. A further breakdown of the dataset, such as for works on historical buildings [47,48] and on improvement and use of the environment [49] could help to clarify the timeline of the public works process. Lastly, the features affecting the mismatch between the expected and actual costs and times to the completion of public works are worth analyzing, also by comparison with existing literature on the topic.

**Author Contributions:** Conceptualization, G.M.; methodology, G.M. and V.A.; formal analysis, G.M. and V.A.; data curation, V.A.; writing—original draft preparation, V.A.; writing—review and editing, G.M. and V.A.; visualization, G. M. and V.A.; supervision, G.M.

**Funding:** This research was funded by the Department of Civil, Architectural and Environmental Engineering of the University of Padova, research grant number 68/2018, prot. n. 560, date 29/03/2018.

**Conflicts of Interest:** The authors have no potential conflict of interest to disclose concerning the research, authorship, and/or publication of this article.

### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
