Next Article in Journal
Structure Formation and Curing Stage of Arbolite–Concrete Composites Based on Iron-Sulfur Binders
Next Article in Special Issue
Railway Infrastructure Upgrade for Freight Transport: Case Study of the Røros Line, Norway
Previous Article in Journal
Graphical Empirical Mode Decomposition–Convolutional Neural Network-Based Expert System for Early Corrosion Detection in Truss-Type Bridges
Previous Article in Special Issue
Quantitative Risk Analysis Framework for Cost and Time Estimation in Road Infrastructure Projects
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Exploring the Relationship Between Project Characteristics and Time–Cost Deviations for Colombian Rural Roads

Department of Civil Engineering, Pontificia Universidad Javeriana, Bogotá 111711, Colombia
*
Author to whom correspondence should be addressed.
Infrastructures 2025, 10(7), 178; https://doi.org/10.3390/infrastructures10070178
Submission received: 22 April 2025 / Revised: 18 June 2025 / Accepted: 19 June 2025 / Published: 9 July 2025

Abstract

Rural road programs are essential for enhancing connectivity in remote areas, yet they frequently encounter schedule delays and budget overruns. This study explores the extent to which specific project characteristics influence these deviations in Colombian rural road contracts. A dataset comprising 229 projects was extracted from the national SECOP open-procurement platform and processed using the CRISP-DM protocol. Following the cleaning and coding of 14 project-level variables, statistical analyses were conducted using Spearman correlations, Kruskal–Wallis tests, and post-hoc Wilcoxon comparisons to identify significant bivariate relations I confirm I confirm I confirm hips. A Random Forest model was subsequently applied to determine the most influential multivariate predictors of cost and time deviations. In parallel, a directed content analysis of contract addenda reclassified 22 recorded deviation descriptors into ten internationally recognized categories of causality, enabling an integrated interpretation of both statistical and documentary evidence. The findings indicate that contract value, geographical region, and contractor configuration are significant determinants of cost and time performance. Additionally, project intensity and discrepancies between awarded and bid values emerged as key contributors to cost escalation. Scope changes and adverse weather conditions together accounted for 76% of all documented deviation triggers, underscoring the relevance of robust front-end planning and climate-risk considerations in rural infrastructure delivery. The findings provide information for stakeholders, policymakers, and professionals who aim to manage the risk of schedule and budget deviations in public infrastructure projects.

1. Introduction

Road infrastructure is an essential component of economic and social development worldwide. It facilitates the efficient transportation of goods and people, improves accessibility to essential services, and promotes regional integration [1]. Road infrastructure accounts for a significant percentage of the Gross Domestic Product (GDP) [2]. In Colombia, this sector is important for the economy in terms of employment and income generation. The impact of roads is equally crucial; the road infrastructure construction sector accounts for approximately 4.5% of the national GDP [3]. Despite the importance of construction projects in the country’s economy, infrastructure projects frequently face time and cost overruns, representing a global challenge [4,5]. This challenge has been widely explored in the literature; however, most studies have relied on stakeholders’ perceptions as their primary source of information. Notably, Asia and Africa have been the focus of most of these studies [6]. Therefore, there is a lack of research that concentrates on Latin American regions, particularly studies based on project data rather than personal opinions.
Many studies have examined the magnitude and frequency of time and cost deviations in infrastructure projects. Regarding delays in road projects, research indicates that approximately 87% of these projects experience delays [7]. A study conducted in Australia analyzed 231 road projects and found that, on average, cost overruns occurred in around 16.30% of them [8]. Other studies focusing on a larger set of infrastructure projects revealed that cost deviations averaged 28% above the initial budget, with larger projects seeing variations that could reach up to 40% [9]. In the United States, it was reported that 55% of infrastructure projects encountered cost overruns [10]. Meanwhile, a study in Norway revealed that cost deviations could be as high as 182.7% for road projects, with an average deviation of 7.8% [11]. More recently, a study of Dutch road projects found that 38% faced cost overruns, with a mean overrun of 16.5% [12]. In Latin America, specifically in Colombia, rural road projects experienced a 19% time overrun and an average cost overrun of 8%. Additionally, 26.92% of the projects reported cost deviations, 23.18% noted time deviations, and 15.33% experienced both deviations simultaneously [13].
Rural road infrastructure projects are susceptible to delays and budget overruns [14]. Unlike large-scale urban or national projects, rural road initiatives often occur in environments with greater logistical, financial, and technical challenges. These issues are further compounded by institutional and political factors, such as the limited administrative capacity of local authorities and the instability associated with electoral cycles [15]. As a result, rural road projects encounter unique challenges during the planning and execution phases [16]. Rural roads are essential for enhancing connectivity in dispersed communities as they facilitate access to vital services like healthcare, education, and markets [17,18]. Consequently, any delays or cost overruns in these projects can lead to inefficient public spending and directly undermine the quality of life for rural populations. Despite the strategic importance of rural road infrastructure, there is a notable gap in the academic literature concerning systematic analysis based on empirical data, particularly in Latin America. Most existing research focuses on urban or national-level infrastructure [19] or relies heavily on stakeholder perceptions [20], which restricts the ability to establish clear causal relationships. By providing quantitative evidence on time and cost deviations, this study seeks to enhance the understanding of the underlying dynamics and support the development of targeted mitigation strategies.
Various global and local studies have identified the causes of time and cost deviations in road infrastructure projects [5,21]. The wide-ranging consequences of the deviations have led researchers to employ various methods to address this global issue across different geographical regions. These methods include analyzing stakeholders’ perceptions through questionnaires regarding the main causes of cost and time overruns [22,23,24] and examining empirical data [8,12,25]. However, the existing literature does not indicate whether there is a direct relationship between specific project characteristics and the causes of these time and cost deviations. Therefore, the research aims of this study were to (1) develop statistical models that identify statistically significant relationships between time and cost deviations and the characteristics of rural road projects and (2) analyze how the causes of project deviations relate to their characteristics throughout the project life cycle. This study contributes to a deeper understanding of schedule and budget deviations affecting rural road infrastructure projects across diverse regions worldwide. Project managers and planners may leverage the findings to design and implement more effective mitigation strategies targeting the various factors that lead to these issues. Thus, this study supports efforts to minimize the risk that such deviations may compromise the continuity, efficiency, and sustainability of infrastructure projects.

2. Literature Background

This chapter reviews the literature on the causes of deviations in time and cost for road infrastructure projects. First, the Colombian rural road construction bidding process is described (see Section 2.1). Second, studies focusing on factors that cause cost deviations are presented (see Section 2.2). Third, studies on time deviations are presented (see Section 2.3).

2.1. Bidding Process for Colombian Rural Road Construction

The bidding process for rural road construction projects in Colombia is governed by the guidelines established by the National Public Procurement Agency, Colombia Compra Eficiente [26]. This process consists of four phases: Initiation, Bid, Award, and Execution. In the Initiation phase, key elements are defined, including the date the process is created, the entity promoting the project, the objectives of the contracted process, and the geographic location. During the Bid phase, the project’s value and the type of bidding process are established. This focuses on public bids, the contracting regime, and the timeline for the process. In the Award phase, the contractor is selected. This entity will execute the project in accordance with the technical specifications. This phase also determines the execution value of the contract, its initial duration, and its scope. The Execution phase involves setting the project’s start and end dates, as well as the initial and final costs. Additionally, this phase includes documentation related to closure minutes and any other justifications required for the project.
According to Colombia Compra Eficiente [26], the main stakeholders involved in public procurement processes include the state entity, which owns the contract and is responsible for structuring the specifications, awarding the contract, and overseeing its execution. Next are the contractors, supervisors, and auditors, who may be designated by the state entity or appointed by third parties. Their role is to monitor the execution of the contract and ensure that the contractor meets its obligations. Lastly, the beneficiaries (namely, the citizens) play an indirect yet significant role by participating in citizen oversight mechanisms. These mechanisms are designed to ensure the transparent use of public resources. Figure 1 illustrates Colombia’s public procurement process structure, highlighting the main activities that must be completed before contract execution begins. Understanding the functions of these actors is crucial for identifying potential critical points in the process where delays and cost overruns are likely to occur, particularly in rural road infrastructure projects.

2.2. Causes of Cost Deviations in Road Infrastructure

Researchers around the world have identified various factors contributing to cost overruns in construction projects, resulting in a substantial body of literature on the topic [5,21]. A 2009 study focused on road construction projects in Zambia [23], highlighting adverse weather conditions, changes in project scope, environmental protection requirements, schedule delays, strikes, technical challenges, inflation, and local government pressures as the primary causes of cost escalation. Another study conducted in Palestine found that the political situation was the most significant reason for delays in road projects [27]. In Italy, a life-cycle analysis of cost overruns in road projects revealed that cost underestimation during the initial phases and opportunistic behavior by contractors (who exploit contract renegotiation opportunities to increase costs) were key factors [28]. Additionally, European research has identified political issues and economic conditions as major determinants of cost deviations in transportation projects, with evidence suggesting that improved regulations and a strong legal framework can help mitigate these cost overruns [29].
A literature review identified several factors contributing to cost overruns, including inaccurate initial estimates, design modifications, issues with permitting and procurement, and contextual factors like corruption and political instability [30]. Specifically, empirical data revealed additional factors impacting costs and timelines for rural road projects in Colombia. The findings indicated that higher values for variables such as budget and project intensity correlate with greater cost and time deviations [13,31]. Furthermore, projects with shorter durations tended to report more significant time overruns. The poorest-performing projects were those initiated in the year when council mayors began their terms, those located in municipalities with more resources, and those awarded through a competitive bidding process. The results also illustrated a relationship between cost performance and time performance [13].

2.3. Causes of Time Deviations in Road Infrastructure

The causes of time deviations in road construction projects have been studied globally over the years. In early studies conducted in the United States, it was found that delays in road projects can be attributed to several factors, including adverse weather conditions, project complexity, a lack of competition among contractors during the bidding process, and significant discrepancies between the bid amount and the engineer’s estimate [32]. In Palestine, the main causes of delay in road construction projects include the political situation, project segmentation, awarding contracts to the lowest-priced bidders, delays in payments by the owner, and equipment shortages [27]. Research conducted in Malawi highlighted additional causes of delay, such as fuel shortages, insufficient cash flow for contractors, challenges related to foreign exchange for importing materials and equipment, slow payment procedures by clients, inadequate equipment, delays in relocating utilities, and shortages of construction materials [22].
Studies across five continents have identified poor planning, insufficient risk management, and political decisions as significant factors contributing to cost overruns and delays in road infrastructure projects [33]. In Egypt, researchers noted that the causes of delays include the owner’s financial problems, shortages of construction equipment and materials, design errors, failures in soil investigations, and poor contractor management [24]. In India, the primary causes of delays in road construction projects are complications in the land acquisition process, the displacement of utilities, difficulties in constructing while managing traffic, inadequate project planning, and frequent design changes [34]. In terrorism-affected regions of Pakistan, additional factors contributing to delays include interference from various stakeholders, threats to security, the forced selection of inexperienced contractors due to tribal pressures, limited financial capacity among contractors, and insufficient construction methods [35].
Table 1 summarizes previous research on cost and time deviations in road infrastructure projects. The symbol “✓” indicates that the referenced source reports the corresponding factor. In contrast, the notation “N/P” denotes that the mentioned factor is not present or not addressed in the cited reference. Most studies on time and cost overruns have been conducted in urban areas or along major highways, particularly in Asia, the Middle East, and Africa. While these studies provide valuable information, they largely rely on qualitative or perception-based methodologies, such as surveys and interviews, and generally lack access to official, public, or audited datasets [36,37,38]. Moreover, the literature does not specifically address the unique characteristics of rural road projects or focus on contexts in Latin America, where operational, institutional, and territorial challenges differ significantly. To fill this gap, the present study makes a novel contribution by focusing exclusively on rural road infrastructure projects in Colombia. It utilizes a public government database that contains empirical information from 229 completed projects. Unlike previous research, this study applies statistical techniques within the CRISP-DM framework, allowing for the quantification of relationships between variables and the prioritization of the main causes of deviations based on objective and replicable data. Additionally, it incorporates underexplored variables in the literature, such as contract modality, funding source, intervention level, and local institutional conditions.

3. Research Method

The research method followed the CRISP-DM (Cross-Industry Standard Process for Data Mining) methodology, which consists of six phases: Business Understanding, Data Understanding, Data Preparation, Modeling, Evaluation, and Deployment. Figure 2 illustrates the sequence of activities associated with each phase of the CRISP-DM process [60].

3.1. Phase 1: Business Understanding

This phase began with examining how the contracting process for rural roads in Colombia operates. It focused on the specific characteristics of the projects available on the public data platform and the reasons for time and cost deviations during their execution. A review of the operations involved in creating, managing, awarding, and closing public bidding processes for rural road construction projects was conducted. Additionally, this phase focused on defining this study’s objectives and analytical scope, as well as understanding the motivations and needs of stakeholders involved in rural road infrastructure projects. In this context, the Business Understanding process extended beyond financial considerations to include a holistic view of various dimensions within the project life cycle. This approach enabled the identification of structural issues related to cost overruns and schedule delays in project execution. One of the steps in this phase involved conducting a preliminary diagnostic to identify potential causes of deviations from both time and cost perspectives. A thorough review of the specialized literature was undertaken, which helped identify common factors influencing delays and cost overruns. These factors included unclear contractual scopes, inadequate resource planning, insufficient supervision mechanisms, and poor identification of technical or geographical risks. Furthermore, a review of the operations involved in creating, managing, awarding, and closing public bidding processes for rural road construction projects was conducted. This review followed the guidelines set forth by the National Public Procurement Agency, “Colombia Compra Eficiente.” This governmental entity aims to improve the efficiency of public procurement processes in Colombia, ensuring that public resources are used transparently and effectively [61]. The analysis acknowledged the impact of institutional and regulatory dynamics, such as bureaucratic procedures, fragmented responsibilities, and contract management practices, which can affect project performance. This initial diagnosis helps to identify key variables related to deviations in rural road construction. The outcome of this stage was the definition of research aims that guide data analysis.

3.2. Phase 2: Understanding Data

This phase involved searching for and collecting the necessary data to address the research aims. Information on rural road projects was gathered from publicly available databases on the government procurement platform. The link to access this public data is available [62], directing users to the open data page. Table 2 presents the filters used to define the data of interest.
An initial database containing information on construction contracts for rural road projects in Colombia was downloaded using these filters. Afterward, the available variables were identified and their nature was classified as numerical or categorical. Additionally, the data quality was assessed by reviewing the units of measurement used for each variable, their meanings, and the consistency of their application across the analyzed contracts. The open-access database used in this study provided access to information on awarded, executed, and closed projects across various sectors, with a specific emphasis on road infrastructure projects relevant to this research. Applying the filters outlined in Table 1 identified an initial pool of 1207 projects. However, 798 projects were excluded after a detailed review of their objectives and contract descriptions revealed inconsistencies with this study’s scope. This exclusion process was conducted manually to ensure the dataset’s relevance, resulting in a final sample of 229 projects suitable for analysis. Once the database was refined, each variable of interest was identified and classified as either categorical or numerical. Categorical (or qualitative) variables were attributes that could not be ordered numerically, while numerical (or quantitative) variables represented measurable or countable values, which could be either discrete or continuous.
A total of fourteen independent variables, along with two dependent variables, were selected for this study, as outlined in Table 3. Each variable was defined based on project characteristics and categorized as either numerical (N) or categorical (C), with “N” indicating numerical variables and “C” indicating categorical ones. To enhance the analysis and interpretation of results, the variables were grouped according to the Project Management Body of Knowledge (PMBOK) framework [63], which describes the life cycle of a construction project in four phases: initiation, planning, execution, and closure/control. This categorization provided a structured approach to analyzing how variables related to each project phase may influence time and cost deviations in rural road infrastructure projects.
This study’s dependent variables included time and cost deviations, represented by Equations (1) and (2) [65].
T i m e   d e v i a t i o n = F i n a l   d e a d l i n e o r i g i n a l   d e a d l i n e   o r i g i n a l   d e a d l i n e
C o s t   d e v i a t i o n = F i n a l   c o s t c o n t r a c t   v a l u e   c o n t r a c t   v a l u e
These results were complemented by identifying the causes of deviation for each project through an examination of the documents published on the public data platform. A detailed review of the contract modifications allowed for the identification of the causes that led to deviations in projects in terms of time and/or cost.

3.3. Phase 3: Data Preparation

This phase focused on a systematic approach to preparing and cleaning the data filtered in the previous stage. First, information from multiple sources was consolidated, including contractual documents, technical progress reports, official databases, and internal records from the entities responsible for executing rural road infrastructure projects. This step involved verifying the consistency of primary data, such as start and end dates, estimated and actual costs, and construction milestones. Additionally, it included standardizing the format of key variables, such as project name, contracting authority, type of work, and cost classification. Next, a thorough assessment of data completeness and integrity was conducted to identify inconsistencies. Simultaneously, strategies were implemented to address duplicated or conflicting data by cross-referencing sources. During this process, priority was given to official documents over informal sources to ensure the robustness, traceability, and reliability of the dataset used for subsequent analyses.
Once the databases were consolidated, strategies for classifying relevant variables were implemented to structure the information for statistical analysis. One of the first actions taken was to standardize measurement units. For instance, all financial data was converted into a single local currency and project durations were transformed from months to days. Additionally, categories of works and costs were grouped according to their construction typology to enable comparisons across heterogeneous projects. Key qualitative variables, such as the type of project contract and the geographic conditions of the construction site, were also coded to facilitate their integration into quantitative models. This step was crucial for the subsequent application of multivariate statistical techniques. The resulting dataset was organized into a structured Microsoft Excel spreadsheet, enabling orderly and accessible data management. Furthermore, a data dictionary was developed, which detailed the definition of each variable, its source, the method used for calculation or imputation, and its analytical relevance in the study context. Finally, an exploratory analysis was conducted using R software version 4.5.1 [66]. This process involved unsupervised learning and identifying patterns and relationships between variables. Descriptive statistics of the variables were conducted alongside visualization tools, such as scatter plots and bar charts, to analyze the distribution and behavior of the selected variables [67].
Unsupervised learning is a set of statistical tools intended for situations where we have only a set of features (X1, X2, …, Xn) measured on (n) observations, without an associated response variable (Y) [68]. The goal was to discover interesting patterns among the independent variables before applying the supervised learning approach. Multiple correspondence analysis (MCA) was implemented to explore relationships. MCA is an extension of correspondence analysis that enables the analysis of complex survey data or datasets with numerous qualitative variables [69]. It helped identify patterns and associations that might not have been immediately apparent, revealing underlying structures in the data, which was crucial for exploratory data analysis. In MCA analysis, the components were based on categorical variables. It was then necessary to partition the numerical predictors into bins or bands. Subsequently, the numerical variables were discretized using k-means clustering, which identified the most intuitive partition. The k-means clustering algorithm enabled the division of input data values into distinct clusters. Therefore, the discretization strategy for the input data was carried out using the maximum and minimum values of the dataset, the calculated cluster centers, and the midpoints between each pair of clusters [67]. MCA allowed for the dimensionality reduction of the data and enabled the visualization of associations between categories, which is useful for understanding data structures. Thus, the set of multiple categorical variables was analyzed to see how they combine and relate.

3.4. Phase 4: Modeling

This phase concentrated on implementing statistical models for bivariate analysis to identify relationships between variables. It involved comparing the characteristics of rural road projects in Colombia with the extent of time and cost deviations. Following this, the analysis also examined the relationship between time deviations and the characteristics of these projects. This study included bivariate comparisons of numerical variables and comparisons involving both numerical and categorical variables.

3.4.1. Comparing Numerical Variables

The first step involved comparing cost and time deviations with the numerical characteristics of rural road projects in Colombia. The objective was to determine the statistical relationship between the dependent variables (cost deviations and time deviations) and the independent variables (characteristics of rural road projects). Since both variables were numerical, Spearman’s correlation coefficient (Spearman’s rho) was utilized after the Anderson–Darling test confirmed that the variables did not follow a normal distribution [70]. In this context, the hypotheses for Spearman’s coefficient were established as follows:
H0. 
There is no correlation between the two variables.
H1. 
There is a significant correlation between the two variables.

3.4.2. Comparing Categorical Variables

The second step involved analyzing a numerical variable (dependent) concerning a categorical variable (independent) using the Kruskal–Wallis test. This test was chosen because the analyzed variables exhibited nonparametric behavior, as confirmed by the results of the Anderson–Darling test, which indicated that the variables did not follow a normal distribution. The Kruskal–Wallis test is a nonparametric statistical method that compares two or more independent groups to determine whether they originate from the same distribution [71]. In this context, the hypotheses for the Kruskal–Wallis test were established as follows:
H0. 
The median deviations across the groups are equal; there are no statistically significant differences between the group medians.
H1. 
At least one group shows a statistically significant deviation in its median compared to the others.
As a complement to those variables where the Kruskal–Wallis test demonstrated significance, the Wilcoxon Mann–Whitney test was used to compare paired data and establish which specific group presented different behavior [70].

3.4.3. Multivariate Analysis

The third step involved multivariate analysis to establish the effect of considering variables that interacted simultaneously [72]. If the dependent variable was numerical, the analyst chose from regression models (as in this case, with time and cost deviations); if it was categorical, the analyst selected from classification models. Multivariate analysis was performed comparing time and cost deviations with the independent variables simultaneously. Random Forest was implemented using a decision tree-based learning method. It handled both numerical and categorical variables simultaneously and ranked them by importance [72]. It generated numerous decision trees, each using a random subset of samples and variables. Key parameters included the number of trees and the random variables per tree. Optimal predictors were identified by running these trees and observing the changes in error [68]. The optimal number of trees was found when adding more trees provided no significant performance gain but increased computational cost. The literature suggests using 64 to 128 trees [73]. Random Forest allowed for a different analysis by considering the interaction between independent variables from a nonlinear perspective and ranking them in order of importance.

3.5. Phase 5: Evaluation

This phase focused on evaluating the relevance and practical significance of the results obtained from the statistical models used. First, the statistical significance of the variables identified as determinants of cost and time deviation was assessed by comparing the findings with empirical evidence and previous studies on infrastructure project management. To achieve this, estimated coefficients and their corresponding confidence intervals were analyzed to ensure that the parameters were statistically significant and aligned with theoretical frameworks and practical knowledge in the field of road and highway construction, particularly in rural areas.

3.6. Phase 6: Deployment

This phase focused on evaluating the relevance and practical significance of the results obtained from the statistical models used. First, the statistical significance of the variables identified as determinants of cost and time deviations was assessed by comparing the findings with those reported in previous studies in the literature. To achieve this, the obtained results and significant variables were analyzed. The outcomes of this phase are further elaborated upon and discussed in the corresponding discussion section of this article (see Section 5).

4. Results

4.1. Exploratory Data Analysis

An exploratory data analysis was conducted to understand the variables in the dataset. Table 4 presents the results of the univariate statistical analysis for the numerical variables. Time deviations in the executed projects ranged from 0% to 510%, with a mean of 29%. In contrast, cost deviations were relatively moderate, with a maximum of 54% and an average of 8.6%. These figures suggest that the projects and their respective addenda remained within the legal limits established by Colombian regulations. Regarding the initial contract values, these ranged from 42.11 to 20,264 CLMMW (Current Legal Monthly Minimum Wage), with an average of 1516 CLMMW. However, the coefficient of variation (CV) of 175% indicated a considerable disparity in project sizes. Furthermore, the CVs for time and cost increases during project execution exceeded 100%, highlighting substantial deviations from the initial estimates in time and cost. Regarding variable award growth, this represented the difference between the project value in the planning stage (bidding cost) and the awarded value to the winning bid as a percentage. Although this variable always had negative or zero values, since a contract could not be awarded for an amount greater than the estimated bid, the variable was presented in absolute value. Most projects were bid on at a value very close to zero; however, the maximum value was 6.55%.
The variable project intensity, which evaluated the contract value divided by the number of days, measured the amount of money invested per day. While similar behavior could be expected for projects of the same type, it was observed that although the average was 12.15, there were values as high as 116.86.
Figure 3a illustrates the number of projects that experienced time and cost deviations compared to those that did not have any deviations or showed both types of deviations. It was found that 48.9% of the projects were executed as planned, meaning that nearly half had no deviations. However, 24% of the projects experienced both time and cost deviations. Additionally, 15.3% faced planning or execution challenges that may have contributed to time deviations, while 11.8% reported cost deviations. Figure 3b illustrates the distribution of the analyzed projects according to the year in which the contract was signed. Notably, all selected projects corresponded to the pre-pandemic period to avoid potential biases associated with the COVID-19 pandemic’s impacts on rural road infrastructure projects. The years with the highest number of analyzed projects were 2017, with 64 projects (27.9%), and 2015, with 63 projects (27.5%). In contrast, 2016 had the lowest number of projects included in the sample, with 34 projects, representing 14.8% of the total analyzed. Figure 3c shows the distribution of analyzed projects based on the promoting entity. A significant concentration was noted among municipal-level promoters, who were responsible for 217 projects (94.8%) of the total projects. In contrast, only 12 projects (5.2%) were initiated by other entities, such as departmental governments or national-level institutions.
Figure 4a summarizes the composition of the sample in terms of project type. New construction contracts accounted for the largest share, with 96 projects (41.9%), highlighting the government’s recent push to expand the tertiary road network rather than simply upgrading existing roads. Improvement works, which typically involve widening, geometric corrections, and drainage upgrades, ranked a close second, accounting for 87 projects (38.0%). Together, these two categories represented nearly four-fifths of the total portfolio, emphasizing a strategic focus on enhancing rural connectivity through capital-intensive initiatives. In contrast, routine maintenance agreements amounted to 37 projects (16.2%), indicating that preventive upkeep receives less budgetary attention than expansion efforts, despite being recognized as cost-effective for maintaining service levels. Rehabilitation projects, aimed at restoring severely deteriorated sections to their original operating standards, were comparatively rare, with only nine cases (3.9%). This limited number may have resulted from successful preventive maintenance in previous years or suggest that major restorative actions are being postponed until roadway conditions become critical.
Figure 4b illustrates the geographic distribution of the 229 rural road projects analyzed. A significant majority of these projects, totaling 186 (81.2%), were located in the Andean region. In contrast, the Orinoquía region accounted for only 18 projects (7.9%), while the Caribbean and Pacific regions had 11 (4.8%) and 10 projects (4.4%), respectively. The Amazon region had the smallest share, with just four initiatives (1.7%). This noticeable imbalance reflects Colombia’s demographic and economic center: most rural settlements, agricultural corridors, and tertiary roads in need of rehabilitation or upgrading are found in the Andean highlands. Conversely, the Amazon and Pacific basins, which are characterized by lower population densities and limited road networks, attract relatively fewer investments. Figure 4c shows the project distribution based on the winning contractors’ legal status. Consortia were the most common, securing 89 contracts (38.9%), which highlights the frequent need for combined technical and financial resources to meet the public tender requirements for road construction projects. Corporate entities that participated individually followed, winning 72 projects (31.4%). Natural persons obtained 58 contracts (25.3%). The “other” category, which included foundations and mixed-economy firms, accounted for only 10 projects (4.4%). The dominance of consortia and corporate entities indicated that institutional clients prefer bidders who can leverage multidisciplinary teams and substantial bonding capacity, attributes that individual contractors often lack for rural infrastructure programs.
Then, unsupervised learning was developed through MCA, including the independent variables: bidding value, initial duration, project intensity, award growth, year, promoter, contractor, project type, and region, to analyze the relationships between them. The process began with the generation of new components. The explanatory capacity of each component was identified and the proportion of variance accounted for by each dimension was determined. The first component explained 10.2% of the variance and the second component explained 8.3% of the variance. One of the inherent characteristics of MCA is that the components, dimensions, or factors created do not necessarily explain large percentages of the total variance [74]. A biplot of the multiple correspondence analysis (MCA) with dimension 1 (Dim1) as the x-axis and dimension 2 (Dim2) as the y-axis helped to identify the variables that were most related to each dimension (see Figure 5). Variables with similar profiles were grouped. It could be observed that the variable “Bidding value” was most related to dimension 1 (x-axis). Since this variable was discretized, the group that contributed the most was the one corresponding to projects with a value greater than 9590 CMMLV (bidding value between 9590 and 20305). Similarly, promoters in the category “Others” also contributed to dimension 1. The variable project intensity contributed significantly (y-axis) to dimension 2, with values ranging from 78.8 to 1217, mainly corresponding to the year 2019.
The ranking of variables contributing, in percentage, to dimension 1 (the first component) is shown in Figure 6. The red dashed line in the above graph indicates the expected average contribution. For a certain component, a variable with a contribution exceeding this threshold is considered important in contributing to the component. Variables above this line were the most important contributors (in this case all the included in the figure). The categories with a bidding value higher than 9590 and a project intensity higher than 50.2 were most important in defining the first dimension and also appeared in the first places in the second dimension. They were the most important for explaining the variability of the dataset.
The MCA plot reveals that certain variables exhibited distinct patterns within specific categories (see Figure 7). It is important to analyze the groups that are clearly separated from the others, such as bidding value for higher amounts (9590.00 to 20305.00 CLMMW) or initial duration for longer projects (492 to 900 days). Similarly, projects with higher project intensity (50.20 to 117.00) and award growth (3.20 to 6.55) display distinct behaviors compared to other groups. Additionally, variation was observed in the behavior of promoters within the “Others” category. However, variables such as year, project type, region, and contractor type did not contribute significantly to the variability in the data and tend to overlap. This analysis allowed for the identification of variables that exerted the most influence on the outcomes and those that did not have a significant impact.

4.2. Analysis Between Numerical Variables

The analysis of numerical variables using bivariate methods started with an Anderson–Darling test to evaluate the normality of each distribution. All variables returned a p-value of less than 0.05, indicating a significant departure from normality. This result justified the use of the non-parametric Spearman rank-order correlation for cost deviation and time deviation.

4.2.1. Cost Deviation Versus Numerical Variables

The bivariate analysis for cost deviation and numerical variables is presented in Table 5, which includes Spearman’s rho and corresponding p-values. Only variables with results that were statistically significant at p-value < 0.05 are included. Cost deviation (CD) showed positive correlations (Spearman’s rho between 0.14 and 0.29) with various planning-stage indicators, including initial contract value, bidding value, award growth, project intensity, and initial contract duration. Larger projects were slightly more susceptible to cost overruns, likely due to their administrative and technical complexity. Additionally, project intensity was a significant variable, measuring the amount of money spent per day on a project. The results showed that higher values of this variable corresponded to greater differences. Project intensity was determined during the planning stages [64]; therefore, errors in estimating it could have been related to failures during this phase. Regarding award growth, the greater the difference between the winning bid and the estimated cost by the owner, the higher the value of cost deviation. A desirable outcome for the owner might have been awarding a contract to a bid that was more economical than the estimated value. Still, the results showed that this could also lead to cost overruns. Therefore, this aspect needs to be carefully reviewed in detail.
During the execution and closure phases of a project, the relationship between CD and project updates became stronger (Spearman’s rho coefficient between 0.32 and 0.99), with the additional cost showing a higher correlation. However, the results highlighted a relationship between cost deviation and time deviation. These findings support the “cost-time feedback loop” commonly observed in infrastructure projects: schedule extensions lead to increased overall costs, adjustments linked to indices, and scope changes, all of which contribute to higher final costs.

4.2.2. Time Deviation Versus Numerical Variables

The bivariate analysis for time deviation and numerical variables is presented in Table 6, which includes Spearman’s rho and corresponding p-values. Only variables for which the result was statistically significant with a p-value << 0.05 are included. Time deviation (TD) was also sensitive to scale effects. The initial contract value, bidding value, and project intensity suggested that larger and more complex projects tended to experience greater delays (Spearman’s rho between 0.28 and 0.36). This may have been due to longer critical paths and a higher exposure to risks, such as financing bottlenecks, stakeholder interference, increased complexity, and the involvement of multiple stakeholders. During the execution phase, TD was closely aligned with additional time and also showed significant correlations with final duration, similar to the previous results (Spearman’s rho between 0.31 and 0.98).

4.3. Analysis of Numerical and Categorical Variables

To explore the relationships between time and cost deviations and categorical variables, the Kruskal–Wallis test was implemented, as the distribution of the numerical variables did not meet the assumptions required for parametric tests (non-normality).

4.3.1. Cost Deviation Versus Categorical Variables

The bivariate analysis of cost deviation and categorical variables is summarized in Table 7, which presents the results of the Kruskal–Wallis test. Only statistically significant variables are included. For these variables, pairwise comparisons were additionally conducted using the Wilcoxon Mann–Whitney test to identify which specific groups showed significant differences. In each case, the category that exhibited different behavior from the others is presented, along with its mean and median values.
The year 2016 stood out from the others due to higher values for cost deviation, and it coincided with the mayoral election year. This supported the hypothesis that electoral turnover disrupts budget control through the renegotiation of priorities and the replacement of supervisory teams. Construction projects differed from the others, with lower values in cost deviation. Maintenance and rehabilitation projects in this case showed higher deviations, probably due to the execution along the roads and their operation, which could be affected by interactions with the community and the inability to work continuously. Regarding the promoter, the executors, which were municipalities, showed a higher cost deviation in the projects they carried out. The others corresponded to community action boards or community associations, in which other factors could influence the outcomes. The management of these projects may differ significantly from that of municipalities, which are governed by different legislation.
The statistical analysis presented above revealed that projects contracted in 2016 had the highest median schedule deviation (39%) and cost deviation (21%) among the entire 2013–2019 panel. A purely descriptive interpretation would label 2016 as an “outlier year,” yet such a label does little to advance the understanding of why time and cost performance deteriorated so sharply. A closer examination of Colombia’s political–institutional landscape revealed that 2016 combined three mutually reinforcing shocks (municipal administrative turnover, national fiscal austerity, and uncertainty surrounding the peace agreement plebiscite) that together disrupted the execution of rural road contracts. Although each factor has been documented separately in the public administration literature, their simultaneous occurrence in a single budget cycle created an especially unfavorable environment for project delivery, thereby offering a substantive causal narrative that complements the purely statistical correlations.
First, Colombia’s local elections were held in late 2015, with newly elected mayors and municipal councils assuming office on 1 January 2016. The first year of a municipal administration invariably entails cabinet reshuffles, renegotiations of advisory contracts, and the drafting and approval of the legally binding Plan de Desarrollo Municipal (MDP). During this transition, ongoing public work contracts are subject to intensified legal and technical review, and supervisory engineers are frequently replaced. Empirical studies of Colombian procurement show that various municipal infrastructure contracts experience at least one suspension or amendment in the first two quarters of a new mayoral term, largely because newly appointed legal teams revisit inherited obligations and adjust investment priorities [13]. Such “administrative friction” slows the certification of completed work and delays cash payments to contractors, a sequence that leads directly to time extensions and cost overruns, as captured in the additional cost variable of our database. In the 2016 subsample, 63% of rural road contracts signed in October–December 2015 were amended between February and August 2016, 21 percentage points higher than the long-term series average, and the mean amendment added 78 calendar days to the contractual schedule.
Second, at the national level, the collapse of oil price revenues in 2015 eroded the central government’s fiscal space and triggered a mid-year expenditure cut in the GDP. While tertiary road projects were predominantly financed with municipal resources, an average of 30% of their budgets depended on co-financing transfers from the Sistema General de Regalías and Invías programs targeted at rural connectivity. Contractors, therefore, faced a liquidity squeeze that forced them to decelerate on-site activities or secure bridge financing at higher interest rates. Both responses had predictable consequences: slower field progress lengthened the critical path, and more expensive working capital raised effective unit costs. Because Colombian public work contracts use indexed escalation formulas only for materials and fuel (not for financial charges), the extra interest payments are typically absorbed as overhead, contributing to the cost deviations recorded in SECOP audit trails.
The third destabilizing element was the political uncertainty surrounding the 2 October 2016 plebiscite on the final peace accord between the Colombian government and the FARC insurgency. The unexpected outcome generated a four- to six-week legislative hiatus while negotiators revised the agreement and Congress awaited clarification on its fiscal implications. During this period, the National Planning Department and the Ministry of Finance issued circulars advising contracting authorities to refrain from launching new procurement processes until the revised accord’s budgetary footprint could be quantified. SECOP records show that the number of new bidding invitations in October and November 2016 decreased by 35% relative to the average of the same months in 2013–2015. Although our sample focused on contracts already underway, the administrative pause had indirect effects: supervising agencies diverted personnel to peace-related tasks, and contractors preparing claims for time extensions or compensation faced delayed responses. Moreover, resources initially earmarked for generic rural development programs were temporarily reprioritized toward Zonas Más Afectadas por el Conflicto (ZOMAC), creating further cash flow uncertainty for projects outside those geographies. Early drafts of the post-conflict implementation decrees also contemplated labor quota requirements for ex-combatants, forcing some contractors to revisit staffing plans and subcontracting chains.

4.3.2. Time Deviation Versus Categorical Variables

The bivariate analysis of time deviation and categorical variables is summarized in Table 8, which presents the results of the Kruskal–Wallis test. In this case, the year 2016 stood out from the others due to higher values for time deviation as it coincided with the mayoral election year, similar to cost deviation. This supported the hypothesis that electoral turnover disrupts budget control through the renegotiation of priorities and supervisory teams. Contractor type was also a significant variable, with Consortium contractors differing from others, with higher values for time deviation. Complex governance arrangements, multi-layered decision chains, and heterogeneity in resource mobilization within consortia can prolong approval cycles for design changes and claim negotiations, thereby extending project schedules.

4.4. Analysis of Cost and Time Deviations and Project Characteristics

This study identified statistically significant relationships between the causes of time and cost deviations and the characteristics of rural road infrastructure projects in Colombia. The analysis included 229 completed projects, and the findings are summarized across several key dimensions. First, the exploratory analysis revealed that time deviations ranged from 0% to 510%, with an average deviation of 29%. In contrast, cost deviations were less pronounced, with a maximum deviation of 54% and an average of 8.6%. This average was within the limits established by Law 80 of 1993, which permits a maximum of 55% for cost additions in publicly funded projects.
Bivariate analysis revealed significant associations between cost and time deviations and variables such as initial contract value, project intensity, bidding value, year, additional cost and time, and initial and final contract values. Only variables such as initial duration and award growth were statistically significant for cost deviations. The analysis revealed that rural road infrastructure projects in Colombia are particularly susceptible to deviations, particularly in terms of cost overruns and delays. These findings highlight the need to improve planning practices and risk management strategies to mitigate the effects of such deviations. A summary of the statistically significant variables is presented in Table 9. Statistical significance is represented by the symbol “✓”.
The findings of this study reveal that rural road infrastructure projects in Colombia are highly exposed to both time and cost overruns. This conclusion is based on a quantitative analysis of 229 completed projects, of which more than 70% experienced delays exceeding three months and approximately two-thirds surpassed their initial contract values by more than 15%. These figures not only indicate a statistically significant recurrence but also suggest the presence of structural patterns that specifically affect this type of intervention. The most frequently identified causes—such as scope changes, scheduling deficiencies, contractor financial issues, delayed payments, and adverse weather conditions—are particularly critical in rural environments, where local institutional capacity is often limited, technical oversight is more challenging, and the margin for managing unforeseen events is substantially reduced. Moreover, the geographic dispersion and accessibility constraints inherent to rural areas increase logistical costs and reduce operational efficiency, further amplifying the impact of any planning or execution discrepancies. Collectively, these factors create a scenario in which rural projects are not only more vulnerable to deviations but also face greater obstacles in correcting them in a timely manner. This highlights the need for tailored management approaches that are specific to the unique conditions of the Colombian rural context.

4.5. Multivariate Analysis of Cost and Time Deviations and Project Characteristics

In the Random Forest model, the following independent variables were included: bidding value, initial duration, project intensity, award growth, year, promoter, project type, contractor type, and region. No numerical variables that might have been correlated with each other were included to avoid noise in the model. First, a comparison of the error reduction versus the number of trees determined an optimal number of trees; in this case, 92 trees were selected. After running the models, an optimal number of five predictors was obtained, based on the reduction in the out-of-bag error. Next, the most important predictors were ranked, considering the increment in Mean Squared Error (MSE) if the variable was eliminated (see Figure 8). In this case, initial duration was followed by bidding value, year, project intensity, and award growth.
Next, the same steps were developed for time deviation and the most important predictors were ranked (see Figure 9). In this case, project intensity was the primary factor, followed by bidding value, year, initial duration, and award growth.
The Random Forest model analysis identified several key variables that contributed to both cost and time deviations. These findings highlight the varying impact of different variables on project deviations. It can be observed that when combining numerical and categorical variables, the numerical variables were the most important and had the most relevance in time and cost deviations.

4.6. Analysis of the Identified Causes and Their Impact on Rural Road Projects

Table 10 illustrates the relationship between the causative factors identified in the international literature and the specific labels found in the contractual addenda and supporting documents of the 229 rural road projects analyzed in this study. A notable observation was the overwhelming prevalence of scope-related drivers, which accounted for 88 entries, or 59% of the total. This trend aligned with global evidence suggesting that poorly defined work packages, design omissions, and overly optimistic front-end planning are the most significant predictors of time and cost overruns in linear infrastructure [75,76]. The high frequency of scope adjustments for Colombian rural roads is not surprising, considering that rights-of-way often pass through geotechnically unstable areas and socially diverse communities. These conditions encourage post-award design optimization. However, the extent of the issue highlighted here reveals an institutional paradox: although the national contracting statute limits additive works to 50% of the original contract amount, project owners frequently implement small modifications that cumulatively approach, though rarely exceed, this statutory threshold.
Table 11 presents the causes analyzed based on the causes identified in the literature. Weather-induced aspects were the second most frequent category, accounting for 25 occurrences, or 17% of the total. This was lower than the frequency of scope changes, but it remained important in operational terms. Unlike megaprojects that typically have detailed hydro-meteorological baselines, contracts for tertiary roads rarely include adequate climate risk allowances. As a result, contractors are left vulnerable to El Niño and La Niña phenomena, which can intensify both the amount and duration of rainfall in Andean watersheds. These findings align with recent studies in Colombia [5,31] that identified weather conditions as a common reason for contract deviations. Changes in the original designs were noted nine times, while social conflicts were mentioned eight times, placing them in the third and fourth positions, respectively. Design adjustments typically indicated technical oversights, whereas social conflicts highlighted the changing governance landscape in post-Agreement rural Colombia. In this context, community expectations regarding labor quotas, environmental protections, and benefit-sharing have increased. Although less frequent, with six or fewer occurrences each, the remaining categories (planning deficiencies, input price volatility, difficult site access, labor scarcity, and financial mismanagement) should not be overlooked. Their combined impact can still undermine performance margins beyond acceptable limits, particularly when they coincide with more common factors.

5. Discussion

In rural road projects in Colombia, a significant 51.1% of the evaluated projects presented some form of deviation, with time deviation being the most frequent issue encountered. This analysis revealed an incidence rate of 15.3% for time deviations, alongside an average magnitude of 29% in time deviations and 8.6% in cost deviations. Previous studies that considered alternative methods of bidding and award reported a noteworthy cost deviation of 19% and an 8.00% time deviation [13]. Typically, higher-value projects are awarded through public bidding processes, which might explain the prevalence of greater values for time deviations. This research found a positive correlation between the size of the contract, measured by its monetary value, and both types of deviations. Other authors have reported cost deviations of around 20% in road projects across various countries, highlighting a global trend in this issue [4].
In specific regions, researchers have reported that the average ratio of planned contract duration to actual completion time is only 58.24%, with observed ranges between 2% and 172% in Saudi Arabia [55]. Regarding the magnitude of cost deviations, an average increase of 18.75% has been documented in road infrastructure projects in India [57]. In Europe, an extensive analysis of 1091 transportation projects in Portugal found an average of 17.8% in cost deviations [77], while a comprehensive study of 620 road projects in Norway reported an average deviation of 7.9%, with values fluctuating between −59% and 183% [11]. Compared to other countries, cost deviations in Colombia are considerably lower. This can be attributed to legal issues surrounding contract management, as deviations exceeding 50% of the initial contract value are not permitted by regulation. However, in terms of time deviations, the figures reported in Colombia are situated in the mid-range when compared to other international contexts.
Regarding significant variables influencing these deviations, this research found a notable relationship between both time and cost deviations, a finding also reported in previous studies [13,32] Additionally, project size has been identified as a key factor influencing both time and cost deviations, with larger projects being consistently associated with higher levels of these deviations, a trend supported by earlier studies [10,28]. For cost deviations specifically, there are additional significant variables at play, including project intensity, which relates to an optimism bias affecting decisions about the funds to be invested per day during the project planning stage, as noted in the previous literature [12]. Another significant factor related to cost deviation is award growth, which pertains directly to the bidding process. An award granted to a contractor with a bid value significantly different from the final contract value has been associated with greater deviations, indicating a potential misalignment in expectations and realities. Creating competitive bidding processes where the bid value receives relevant scoring can inadvertently be detrimental to project outcomes. Previous studies have also reported errors in the bidding and award processes as significant contributing factors to these deviations [28,78].
Using comprehensive data derived from settled and stored projects in public databases lays a solid foundation for future research to investigate these relationships in greater depth, both within the Colombian context and in other countries. This ongoing research effort aims to validate the findings presented herein and open new discussions on the similarities and differences that may arise in global comparative studies. Such discussions are crucial for fostering a deeper understanding of the factors that influence project success and developing more effective management strategies for road infrastructure projects worldwide.
This study employed a cross-sectional correlational design, which means that the observed relationships between project characteristics and time and cost overruns reflect statistically significant associations but do not permit definitive causal inferences. The absence of comprehensive control variables and the lack of longitudinal data limit the ability to isolate the independent effect of each factor and control for potential confounders. Nevertheless, given the exploratory nature of the analysis and the understudied context of rural roads in Colombia, the approach adopted was appropriate for identifying relevant and priority patterns that can guide future research and inform management decisions. The government database used in this study is extensive, audited, and constitutes a reliable source that enhances the external validity of the results. It is recommended that future research incorporate longitudinal designs, additional control variables, and advanced causal inference techniques to improve the understanding of the mechanisms driving cost and time deviations.

6. Conclusions

This study identified statistically significant variables that influence time and cost deviations in rural road projects in Colombia based on empirical data obtained from the Colombian Public Electronic Procurement System (SECOP). This research established a statistically significant relationship between time and cost deviations in rural road projects in Colombia. The findings indicate that when projects experience delays, they are also likely to encounter cost overruns. This correlation suggests that factors contributing to time deviations can simultaneously impact project budgets, highlighting the interconnected nature of these two types of deviations. Understanding this relationship is crucial for project managers and stakeholders to implement effective strategies to mitigate both time and cost overruns, ultimately leading to more successful project outcomes. Among the variables analyzed, both time and cost deviations exhibited a positive correlation with certain key factors. Specifically, the contract value and the estimated cost established during the bidding process were found to significantly influence these deviations. Projects with higher contract values tended to experience greater time and cost overruns, suggesting that the scale and complexity associated with larger contracts may contribute to increased challenges in project execution. Similarly, discrepancies between estimated costs and actual expenditures further underscore the importance of accurate budgeting and forecasting in mitigating potential deviations.
This analysis identified significant variables contributing to cost deviations in rural road projects. Notably, project intensity emerged as a critical factor, indicating that the level of activity and resource deployment can influence overall project costs. Additionally, award growth, which relates to the planning stage and bidding process, was also identified as a significant variable. This suggests that discrepancies between the awarded bid and the original contract value can lead to heightened cost overruns. Understanding these variables is crucial for stakeholders to develop effective planning and bidding strategies that help mitigate potential cost deviations and improve project performance. This research indicates that the years in which mayors are elected tend to report higher cost deviations in public projects. This phenomenon suggests that political factors, including leadership changes and associated shifts in project priorities, may significantly influence project budgets. Understanding the impact of these political cycles is crucial for stakeholders to devise strategies that mitigate the adverse effects of political factors on project delivery and budgeting.
Project type and the involved promoter significantly impact cost deviations. Maintenance projects, in particular, report higher deviations. The experience and approach of the project promoter are crucial. Acknowledging the influence of project type and promoter capabilities is crucial for developing effective strategies to minimize cost deviations and enhance project outcomes. Consortium contractors tend to exhibit higher time deviations, likely due to their involvement in larger-scale projects with higher overall values. The complexity and intricacies of managing substantial projects can lead to unforeseen challenges, resulting in increased delays. Additionally, consortia often comprise multiple firms, which can create coordination issues and complicate decision-making processes, further exacerbating timelines. These factors, when combined, can contribute to significant schedule overruns as the scale and scope of the projects expand. Thus, understanding the dynamics of consortium contracting is crucial for stakeholders aiming to mitigate time deviations and improve the timely completion of high-value projects.
The results of this study expand our understanding of how the specific characteristics of projects impact their performance, providing not only empirical evidence for future studies but also a replicable contextual framework for planning and managing rural infrastructure projects in Colombia. The results allow project managers to anticipate and mitigate deviations. This study is limited by the exclusive focus on rural road projects and a specific period (2015–2019). Expanding the database to include urban infrastructure projects and exploring broader temporal patterns are suggested. Furthermore, future research could utilize nonlinear predictive models, such as neural networks, to analyze complex causes and propose solutions tailored to each specific context.
Building on the results provided by the Random Forest analysis, future work could implement additional machine learning strategies that are capable of capturing higher-order, non-linear interactions among project variables while maintaining model interpretability. First, a feed-forward neural network (multilayer perceptron) could be trained using the full SECOP dataset, with a focus on project phase-specific feature blocks. Hyperparameters, such as the number of hidden layers, neurons, and dropout rates, could be optimized through a Bayesian search within a nested five-fold cross-validation scheme to prevent overfitting. The resulting weights could be analyzed using SHAP values to rank each predictor’s marginal contribution and visualize interaction effects that conventional statistics might overlook. Second, gradient-boosted decision-tree ensembles, such as XGBoost, could be benchmarked against the neural model and the baseline Random Forest. This approach would aim to achieve competitive error reduction while providing transparent, rule-based explanations derived from tree paths. Third, a support vector machine with a radial basis kernel could be calibrated on a SMOTE-balanced subsample to assess its robustness in scenarios with pronounced class imbalance, such as extreme overruns versus nominal compliance. The distances to the hyperplane could allow for the clear identification of contracts that lie near the decision boundary, warranting early managerial attention. The complementary explanatory artifacts generated by these models could be integrated into an actionable dashboard for procurement agencies. This could enable them to forecast deviation risk at the tender stage and design preemptive interventions.

Author Contributions

Conceptualization, J.Q., A.M., A.G.-C., and O.S.; methodology, J.Q., A.M., A.G.-C., and O.S.; software, J.Q., A.M., and A.G.-C.; validation, J.Q., A.M., and A.G.-C.; formal analysis, J.Q., A.M., A.G.-C., and O.S.; investigation, J.Q., A.M., A.G.-C., and O.S.; resources, A.G.-C. and O.S.; data curation, J.Q., A.M., and A.G.-C.; writing—original draft preparation, J.Q., A.M., A.G.-C., and O.S.; writing—review and editing, J.Q., A.M., A.G.-C., and O.S.; visualization, J.Q., A.M., A.G.-C., and O.S.; supervision, A.G.-C. and O.S.; project administration, A.G.-C.; funding acquisition, A.G.-C. and O.S. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the Pontificia Universidad Javeriana, Colombia, through the call “Apoyo a proyectos de investigación liderados por profesores que se encuentran en su primera etapa 2023” with the project entitled “Propuesta de buenas prácticas para procesos de contratación de proyectos construcción de vías rurales en Colombia a partir de datos públicos (ID 20699)”. Similarly, the APC was funded by the Pontificia Universidad Javeriana, Colombia.

Data Availability Statement

The original contributions presented in this study are included in the article. Further inquiries can be directed to the corresponding author(s).

Acknowledgments

Omar Sánchez thanks Minciencias (former Colciencias) for the sponsorship and support through the “Convocatoria Doctorados Nacionales—2015” program. Minciencias is the Science, Technology, and Innovation Ministry, a Colombian government agency that supports fundamental and applied research in the country.

Conflicts of Interest

The authors declare no conflicts of interest.

Declaration of Generative AI and AI-Assisted Technologies in the Writing Process

During the preparation of this work, the authors used Grammarly in order to improve the readability, spelling, and grammar. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of this publication.

References

  1. Jedwab, R.; Moradi, A. The Permanent Effects of Transportation Revolutions in Poor Countries: Evidence from Africa. Rev. Econ. Stat. 2016, 98, 268–284. [Google Scholar] [CrossRef]
  2. Asiedu, R.O.; Adaku, E. Cost Overruns of Public Sector Construction Projects: A Developing Country Perspective. Int. J. Manag. Proj. Bus. 2019, 13, 66–84. [Google Scholar] [CrossRef]
  3. DANE Departamento Administrativo Nacional de Estadística-DANE. Available online: https://www.dane.gov.co/index.php/estadisticas-por-tema/cuentas-nacionales/cuentas-nacionales-trimestrales/pib-informacion-tecnica (accessed on 19 March 2025).
  4. Flyvbjerg, B.; Skamris Holm, M.K.; Buhl, S.L. What Causes Cost Overrun in Transport Infrastructure Projects? Transp. Rev. 2004, 24, 3–18. [Google Scholar] [CrossRef]
  5. Mejía, G.; Sánchez, O.; Castañeda, K.; Pellicer, E. Stakeholders’ Issues as a Source of Project Delays: A Meta-Analy-Sis between Building and Road Projects. Rev. Construcción. 2023, 22, 51–73. [Google Scholar] [CrossRef]
  6. Gómez-Cabrera, A.; Gutierrez-Bucheli, L.; Muñoz, S. Causes of Time and Cost Overruns in Construction Projects: A Scoping Review. Int. J. Constr. Manag. 2023, 24, 1107–1125. [Google Scholar] [CrossRef]
  7. Anastasopoulos, P.C.; Labi, S.; Bhargava, A.; Mannering, F.L. Empirical Assessment of the Likelihood and Duration of Highway Project Time Delays. J. Constr. Eng. Manag. 2012, 138, 390–398. [Google Scholar] [CrossRef]
  8. Creedy, G.D.; Skitmore, M.; Wong, J.K.W. Evaluation of Risk Factors Leading to Cost Overrun in Delivery of Highway Construction Projects. J. Constr. Eng. Manag. 2010, 136, 528–537. [Google Scholar] [CrossRef]
  9. Venkateswaran, C.; Murugasan, R. Time Delay and Cost Overrun of Road over Bridge (ROB) Construction Projects in India. J. Constr. Dev. Ctries. 2017, 22, 79–96. [Google Scholar] [CrossRef]
  10. Bordat, C.; McCullouch, B.; Sinha, K. An Analysis of Cost Overruns and Time Delays of INDOT Projects; Joint Transportation Research Program: West Lafayette, Indiana, 2004. [Google Scholar]
  11. Odeck, J. Cost Overruns in Road Construction—What Are Their Sizesand Determinants? Transp. Policy 2004, 11, 43–53. [Google Scholar] [CrossRef]
  12. Cantarelli, C.C.; Molin, E.J.E.; van Wee, B.; Flyvbjerg, B. Characteristics of Cost Overruns for Dutch Transport Infrastructure Projects and the Importance of the Decision to Build and Project Phases. Transp. Policy 2012, 22, 49–56. [Google Scholar] [CrossRef]
  13. Gómez-Cabrera, A.; Sanz-Benlloch, A.; Montalban-Domingo, L.; Ponz-Tienda, J.L.; Pellicer, E. Identification of Factors Affecting the Performance of Rural Road Projects in Colombia. Sustainability 2020, 12, 7377. [Google Scholar] [CrossRef]
  14. Alashwal, N.; Alashwal, A. Rural Access Programs in Yemen: Delay of International Funded Road Projects. Eng. Constr. Archit. Manag. 2023, 30, 1746–1767. [Google Scholar] [CrossRef]
  15. Abdul Latiff, A.M.; Jaapar, A.; Mat Isa, C.M. Factors Influencing Governance Practices in Rural Public Infrastructure Projects: A Case Study of Rural Road Project in Malaysia. Int. J. Sustain. Constr. Eng. Technol. 2021, 12, 35–45. [Google Scholar] [CrossRef]
  16. Musleh Uddin Hasan, M.; Abdul Quium, A.S.M.; Rahman, M.; Khatun, F.; Akther, M.S.; Haque, A.; Jahan, S.; Islam, I.; Ahmed, T.; Shubho, T.H. A Methodology for Planning and Prioritisation of Rural Roads in Bangladesh. Sustainability 2022, 14, 2337. [Google Scholar] [CrossRef]
  17. Lokesha, M.N.; Mahesha, M. Impact of Road Infrastructure on Agricultural Development and Rural Road Infrastructure Development Programmes in India. Int. J. Humanit. Soc. Sci. Invent. 2016, 5, 1–7. [Google Scholar]
  18. Zhou, Z.; Duan, J.; Li, W.; Geng, S. Can Rural Road Construction Promote the Sustainable Development of Regional Agriculture in China? Sustainability 2021, 13, 10882. [Google Scholar] [CrossRef]
  19. Heravi, G.; Mohammadian, M. Investigating Cost Overruns and Delay in Urban Construction Projects in Iran. Int. J. Constr. Manag. 2021, 21, 958–968. [Google Scholar] [CrossRef]
  20. Amoatey, C.; Okanta, A. Exploring Critical Road Project Delay Factors in Ghana. J. Facil. Manag. 2017, 15, 110–127. [Google Scholar] [CrossRef]
  21. Mejía, G.; Sánchez, O.; Castañeda, K.; Pellicer, E. Delay Causes in Road Infrastructure Projects in Developing Countries. Rev. Constr. 2020, 10, 221–235. [Google Scholar] [CrossRef]
  22. Mwamvani, H.D.J.; Amoah, C.; Ayesu-Koranteng, E. Causes of Road Projects’ Delays: A Case of Blantyre, Malawi. Built Environ. Proj. Asset Manag. 2022, 12, 293–308. [Google Scholar] [CrossRef]
  23. Kaliba, C.; Muya, M.; Mumba, K. Cost Escalation and Schedule Delays in Road Construction Projects in Zambia. Int. J. Proj. Manag. 2009, 27, 522–531. [Google Scholar] [CrossRef]
  24. Santoso, D.; Soeng, S. Analyzing Delays of Road Construction Projects in Cambodia: Causes and Effects. J. Manag. Eng. 2016, 32, 05016020. [Google Scholar] [CrossRef]
  25. Catalão, F.P.; Cruz, C.O.; Sarmento, J.M. Public Management and Cost Overruns in Public Projects. Int. Public Manag. J. 2022, 25, 677–703. [Google Scholar] [CrossRef]
  26. Colombia Compra Eficiente Ciclo de Vida de Los Proyectos de Licitación Publica; Agencia Nacional de Contratación Pública: Bogotá, Colombia, 2014.
  27. Mahamid, I. Frequency of Time Overrun Causes in Road Construction in Palestine: Contractors’ View. Organ. Technol. Manag. Constr. 2013, 5, 720–729. [Google Scholar] [CrossRef]
  28. Park, Y.I.; Papadopoulou, T.C. Causes of Cost Overruns in Transport Infrastructure Projects in Asia: Their Significance and Relationship with Project Size. Built Environ. Proj. Asset Manag. 2012, 2, 195–216. [Google Scholar] [CrossRef]
  29. Catalão, F.P.; Cruz, C.O.; Sarmento, J.M. The Determinants of Cost Deviations and Overruns in Transport Projects, an Endogenous Models Approach. Transp. Policy 2019, 74, 224–238. [Google Scholar] [CrossRef]
  30. Herrera, R.F.; Sánchez, O.; Castañeda, K.; Porras, H. Cost Overrun Causative Factors in Road Infrastructure Projects: A Frequency and Importance Analysis. Appl. Sci. 2020, 10, 5506. [Google Scholar] [CrossRef]
  31. Shamdasani, Y. Rural Road Infrastructure & Agricultural Production: Evidence from India. J. Dev. Econ. 2021, 152, 102686. [Google Scholar] [CrossRef]
  32. Bhargava, A.; Anastasopoulos, P.C.; Labi, S.; Sinha, K.C.; Mannering, F.L. Three-Stage Least-Squares Analysis of Time and Cost Overruns in Construction Contracts. J. Constr. Eng. Manag. 2010, 136, 1207–1218. [Google Scholar] [CrossRef]
  33. Flyvbjerg, B. What You Should Know about Megaprojects and Why: An Overview. Proj. Manag. J. 2014, 45, 6–19. [Google Scholar] [CrossRef]
  34. Sharma, V.K.; Gupta, P.K.; Khitoliya, R.K. Analysis of Highway Construction Project Time Overruns Using Survey Approach. Arab. J. Sci. Eng. 2021, 46, 4353–4367. [Google Scholar] [CrossRef]
  35. Zafar, I.; Wuni, I.Y.; Shen, G.Q.P.; Ahmed, S.; Yousaf, T. A Fuzzy Synthetic Evaluation Analysis of Time Overrun Risk Factors in Highway Projects of Terrorism-Affected Countries: The Case of Pakistan. Int. J. Constr. Manag. 2022, 22, 732–750. [Google Scholar] [CrossRef]
  36. Elawi, G.S.A.; Algahtany, M.; Kashiwagi, D.; Sullivan, K. Major Factors Causing Construction Delays in Mecca. J. Adv. Perform. Inf. Value 2015, 7, 75. [Google Scholar] [CrossRef]
  37. Meeampol, S.; Ogunlan, S.O. Factors Affecting Cost and Time Performance on Highway Construction Projects: Evidence from Thailand. J. Financ. Manag. Prop. Constr. 2006, 11, 3–20. [Google Scholar] [CrossRef]
  38. Indhu, B.; Yogeswari, K.; Dhivya, D. Ranking of Delay Factors in Road Construction Project Due to Improper Construction Equipment Management. Int. J. Recent Technol. Eng. 2019, 8, 1962–1971. [Google Scholar] [CrossRef]
  39. Mansfield, N.; Ugwu, O.; Doran, T. Causes of Delay and Cost Overruns in Nigerian Construction Projects. Int. J. Proj. Manag. 1994, 12, 254–260. [Google Scholar] [CrossRef]
  40. Omoregie, A.; Radford, D. Infrastructure Delays and Cost Escalation: Causes and Effects in Nigeria. In Proceedings of the 6th International Postgraduate Research Conference in the Built and Human Environment, Delft, Netherlands, 6–7 April 2006; pp. 79–93. [Google Scholar]
  41. Nasir, A.R.; Gabriel, H.F.; Choudhry, R.M. Cost and Time Overruns in Highway Projects in Pakistan. In Proceedings of the Sixth International Conference on Construction in the 21st Century (CITC-VI) “Construction Challenges in the New Decade”, Kuala Lumpur, Malaysia, 5 July 2011; pp. 1–8. [Google Scholar]
  42. Mahamid, I.; Bruland, A. Cost Deviation in Road Construction Projects: The Case of Palestine. Australas. J. Constr. Econ. Build. 2012, 12, 58–71. [Google Scholar] [CrossRef]
  43. Patil, S.; Gupta, A.; Desa, D.; Sajane, A. Causes of Delay in Indian Transportation Infrastructure Projects. Int. J. Res. Eng. Technol. 2013, 2, 71–80. [Google Scholar]
  44. Kamanga, M.; Steyn, W. Causes of Delay in Road Construction Projects in Malawi. J. S. Afr. Inst. Civ. Eng. 2013, 55, 79–85. [Google Scholar]
  45. Rwakarehe, E.E.; Mfinanga, D.A. Effect of Inadequate Design on Cost and Time Overrun of Road Construction Projects in Tanzania. J. Constr. Eng. Proj. Manag. 2014, 4, 15–28. [Google Scholar] [CrossRef]
  46. Kaleem, S.; Irfan, M.; Gabriel, H.F. Estimation of Highway Project Duration at the Planning Stage and Analysis of Risk Factors Leading to Time Overrun. In Proceedings of the T and DI Congress 2014: Planes, Trains, and Automobiles—Proceedings of the 2nd Transportation and Development Institute Congress, Orlando, FL, USA, 8–11 June 2014. [Google Scholar]
  47. Hasan, R.; Suliman, S.; Malki, Y. An Investigation into the Delays in Road Projects in Bahrain. Int. J. Res. Eng. Sci. 2014, 2, 38–47. [Google Scholar]
  48. Emam, H.; Farrell, P.; Abdelaal, M. Causes of Delay on Infrastructure Projects in Qatar. In Proceedings of the 31st Annual ARCOM Conference; Association of Researchers in Construction Management, Linconln, UK, 7–9 September 2015; pp. 773–782. [Google Scholar]
  49. Honrao, M.Y.; Desai, P.D.B. Study of Delay in Execution of Infrastructure Projects–Highway Construction. Int. J. Sci. Res. Publ. 2015, 5, 1–8. [Google Scholar]
  50. Aziz, R.F.; Abdel-Hakam, A.A. Exploring Delay Causes of Road Construction Projects in Egypt. Alex. Eng. J. 2016, 55, 1515–1539. [Google Scholar] [CrossRef]
  51. Mohamad, S.F.W.; Singh, B. Delays in Road Construction Projects in the State of Perak. Constr. Manag. Transp. 2016, 2, 53–66. [Google Scholar]
  52. Gituro, W.; Mwawasi, S. Time and Cost Overruns in Road Construction Projects in Kenya under Kenya National Highways Authority. Oper. Res. Soc. East Afr. J. 2016, 6, 117–156. [Google Scholar]
  53. Khan, M.F.; Khan, A.; Khan, M.M.; Zeb, F.; Safdar, M.Z.; Mustafa, K.; Shahzada, N. Significant Delaying Factors Affecting Time Overrun in Road Construction Projects in Abbottabad (Pakistan): Contractors Perspective. Abasyn J. Soc. Sci. 2016, 9, 220–230. [Google Scholar]
  54. Amare, Y.; Quezon, E.T.; Busier, M. Causes of Delays during Construction Phase of Road Projects Due to the Failures of Contractor, Consultant, and Employer in Addis Ababa City Road Authority. Int. J. Sci. Eng. Res. 2017, 8, 15–25. [Google Scholar]
  55. Mahamid, I. Schedule Delay in Saudi Arabia Road Construction Projects: Size, Estimate, Determinants and Effects. Int. J. Archit. Eng. Constr. 2017, 6, 51–58. [Google Scholar] [CrossRef]
  56. El-Maaty, A.E.A.; El-Kholy, A.M.; Akal, A.Y. Modeling Schedule Overrun and Cost Escalation Percentages of Highway Projects Using Fuzzy Approach. Eng. Constr. Archit. Manag. 2017, 24, 809–827. [Google Scholar] [CrossRef]
  57. Pai, S.; Patnaik, B.; Mittal, A.; Anand, N. Identification of Risks Causing Time and Cost Overrun in Roads and Highway Projects in India. Int. J. Civ. Eng. Technol. 2018, 9, 683–697. [Google Scholar]
  58. Sánchez, O.G.; Castañeda, K.M.; Herrera, R.F.; Pellicer, E. Benefits of Last Planner® System in Mitigation of Delay Causes in Road Infrastructure Projects. In Proceedings of the XI Simpósio Brasileiro De Gestão E Economia Da Construção, Londrina, Paraná, Brasil, 23–25 October 2019. [Google Scholar]
  59. Al Hosani, I.I.A.; Dweiri, F.T.; Ojiako, U. A Study of Cost Overruns in Complex Multi-Stakeholder Road Projects in the United Arab Emirates. Int. J. Syst. Assur. Eng. Manag. 2020, 11, 1250–1259. [Google Scholar] [CrossRef]
  60. Chapman, P. SPSS, CRISP-DM 1.0: Step-by-Step Data Mining Guide; IBM: Armonk, NY, USA, 2000. [Google Scholar]
  61. Colombia Compra Eficiente Guía Rápida Para Hacer Un Proceso de Contratación de Licitación Pública en el Secop II; Colombia Compra Eficiente: Cundinamarca, Colombia, 2017.
  62. Datos Abiertos Colombia. Available online: https://www.datos.gov.co/ (accessed on 14 April 2025).
  63. Project Management Institute A Guide to the Project Management Body of Knowledge (PMBOK®); Project Management Institute: Newtown Square, PA, USA, 2021.
  64. Federal Highway Administration Alternative Contracting Method Performance in US Highway Construction; Federal Highway Administration: Washington, DC, USA, 2018.
  65. Gransberg, D.D.; Villarreal-Buitrago, M. Construction Project Performance Metrics. In Proceedings of the Annual Conference of AACE International, San Antonio, TX, USA, 23–26 July 2002. [Google Scholar]
  66. R Core Team. The R Project for Statistical Computing. Available online: https://www.r-project.org/ (accessed on 23 July 2024).
  67. Larose, D.T.; Larose, C.D. Data Mining and Predictive Analytics; Wiley Series on Methods and Applications in Data Mining; Wiley: New York, NU, USA, 2015; p. 794. [Google Scholar]
  68. James, G.; Witten, D.; Hastie, T.; Tibshirani, R. An Introduction to Statistical Learning; Springer: New York, NY, USA, 2013. [Google Scholar]
  69. Greenacre, M. Correspondence Analysis in Practice; Chapman and Hall/CRC: Boca Raton, FL, USA, 2007; ISBN 9780429146145. [Google Scholar]
  70. Goos, P.; Meintrup, D. Statistics with JMP; Wiley: Hoboken, NJ, USA, 2016. [Google Scholar]
  71. Agresti, A.; Finlay, B. Statistical Methods for the Social Sciences, 5th ed.; Pearson: London, UK, 2017. [Google Scholar]
  72. Breiman, L. Random Forests. Mach. Learn. 2001, 45, 5–32. [Google Scholar] [CrossRef]
  73. Oshiro, T.M.; Perez, P.S.; Baranauskas, J.A. How Many Trees in a Random Forest? In Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Springer: New York, NY, USA, 2012; Volume 7376, pp. 154–168. [Google Scholar] [CrossRef]
  74. Diaz Monroy, L.G.; Morales Rivera, M.A. Análisis Estadístico de Datos Multivariados; Universidad Nacional de Colombia, Ed.; Universidad Nacional de Colombia: Bogotá, Colombia, 2009; ISBN 9789587751062. [Google Scholar]
  75. Castañeda, K.; Sánchez, O.; Peña, C.A.; Herrera, R.F.; Mejía, G. Comprehensive Analysis of Scheduling Processes in Road Intersections: Associating Deficiencies and Their Consequences in Colombian Projects. Appl. Sci. 2025, 15, 4614. [Google Scholar] [CrossRef]
  76. Flyvbjerg, B.; Holm, M.S.; Buhl, S. Underestimating Costs in Public Works Projects: Error or Lie? J. Am. Plan. Assoc. 2002, 68, 279–295. [Google Scholar] [CrossRef]
  77. Pinheiro Catalão, F.; Cruz, C.O.; Miranda Sarmento, J. Exogenous Determinants of Cost Deviations and Overruns in Local Infrastructure Projects. Constr. Manag. Econ. 2019, 37, 697–711. [Google Scholar] [CrossRef]
  78. Mahamid, I.; Bruland, A.; Dmaidi, N. Causes of Delay in Road Construction Projects. J. Manag. Eng. 2012, 28, 300–310. [Google Scholar] [CrossRef]
Figure 1. Public procurement system in Colombia.
Figure 1. Public procurement system in Colombia.
Infrastructures 10 00178 g001
Figure 2. Research method stages.
Figure 2. Research method stages.
Infrastructures 10 00178 g002
Figure 3. Number of projects according to (a) time and cost deviations, (b) contract signing year, and (c) promoter.
Figure 3. Number of projects according to (a) time and cost deviations, (b) contract signing year, and (c) promoter.
Infrastructures 10 00178 g003
Figure 4. Number of projects according to (a) type of project, (b) region, and (c) type of contractor.
Figure 4. Number of projects according to (a) type of project, (b) region, and (c) type of contractor.
Infrastructures 10 00178 g004
Figure 5. Multiple correspondence analysis (MCA) biplot with dimension 1 (Dim1) as x-axis and dimension 2 (Dim2) as y-axis.
Figure 5. Multiple correspondence analysis (MCA) biplot with dimension 1 (Dim1) as x-axis and dimension 2 (Dim2) as y-axis.
Infrastructures 10 00178 g005
Figure 6. The top 10 variables that contributed the most to the dimensions.
Figure 6. The top 10 variables that contributed the most to the dimensions.
Infrastructures 10 00178 g006
Figure 7. Multiple correspondence analysis (MCA) plot.
Figure 7. Multiple correspondence analysis (MCA) plot.
Infrastructures 10 00178 g007
Figure 8. Variable importance plot based on MSE increase for the first random forest model.
Figure 8. Variable importance plot based on MSE increase for the first random forest model.
Infrastructures 10 00178 g008
Figure 9. Variable importance plot based on MSE increase for the time deviation random forest model.
Figure 9. Variable importance plot based on MSE increase for the time deviation random forest model.
Infrastructures 10 00178 g009
Table 1. Literature review of factors of cost and time deviations in road projects.
Table 1. Literature review of factors of cost and time deviations in road projects.
ID.ReferenceYearCountryRegionData SourceNumber of ProjectsIdentified Factors *
F1F2F3F4F5F6F7F8F9F10F11
1Mansfield et al. [39]1994NigeriaAfricaSurvey + secondary dataN/PN/PN/PN/P
2Meeampol and Ogunlana [37]2006ThailandAsiaSurvey of managers99N/PN/PN/PN/P
3Omoregie and Radford [40]2006NigeriaAfricaOfficial data + surveyN/PN/PN/PN/P
4Nasir et al. [41]2011PakistanAsiaNHA project dataN/PN/PN/PN/P
5Mahamid and Bruland [42]2012PalestineAsiaOfficial database169N/PN/PN/PN/PN/PN/PN/P
6Patil et al. [43]2013IndiaAsiaSurveyN/PN/PN/PN/PN/P
7Kamanga and Steyn [44]2013MalawiAfricaSurveyN/PN/PN/PN/P
8Rwakarehe and Mfinanga [45]2014TanzaniaAfricaOfficial database7N/PN/PN/PN/PN/PN/P
9Kaleem et al. [46]2014PakistanAsiaNHA project dataN/PN/PN/PN/P
10Hasan et al. [47]2014BahrainAsiaSurveyN/PN/PN/PN/P
11Emam et al. [48]2015QatarAsiaInterview + surveyN/PN/PN/PN/PN/P
12Honrao and Desai [49]2015IndiaAsiaSurveyN/PN/PN/PN/P
13Aziz and Abdel-Hakam [50]2016EgyptAfricaSurveyN/PN/PN/PN/PN/PN/PN/PN/P
14Mohamad and Singh [51]2016MalaysiaAsiaSurvey10N/PN/PN/P
15Gituro and Mwawasi [52]2016KenyaAfricaOfficial database24N/PN/P
16Khan et al. [53]2016PakistanAsiaSurveyN/PN/PN/PN/P
17Amare et al. [54]2017EthiopiaAfricaSurvey15N/PN/P
18Mahamid [55]2017Saudi ArabiaAsiaSurvey55N/PN/PN/P
19El-Maaty et al. [56]2017EgyptAfricaSurvey111N/PN/PN/P
20Catalão et al. [29]2018PortugalEuropeOfficial database1091N/PN/PN/PN/PN/PN/PN/PN/P
22Pai et al. [57]2018IndiaAsiaSurveyN/PN/PN/PN/P
23Sanchez et al. [58]2019VariousAmericaOfficial database10N/PN/PN/PN/PN/P
24Al Hosani et al. [59]2020United Arab EmiratesAsiaSurveyN/PN/PN/PN/PN/P
25This study2025ColombiaAmericaOfficial database229N/PN/PN/PN/P
* F1: scope changes; F2: weather conditions; F3: poor financial management; F4 delays in contractor payments; F5 poor planning and scheduling; F6: changes in original designs; F7: contractor’s financial problems; F8: fluctuations in material prices; F9: labor absenteeism; F10: poor financial management by the contracting entity; F11: social conflicts.
Table 2. Filters to delimit the rural road contracts of interest.
Table 2. Filters to delimit the rural road contracts of interest.
IDFilterDescription
F1FamilyThis category includes contracts based on the goods or services involved in the contracting process, considering their primary characteristics as defined by the Colombian Compra Eficiente goods and services classifier. In this study, the category for roads was selected.
F2Project statusThis filter identifies the contract status, classifying it as executed, completed, suspended, or at a specific stage based on the information provided by the entities since the publication date.
F3Contract signing yearThis filter selects contracts according to the year they were signed.
Table 3. Description of independent variables.
Table 3. Description of independent variables.
Project PhaseIDVariableTypeDefinitionUnits/Values
Initiation (4)V1Project TypeCDescribes the nature of the project based on its primary objective.Construction, Improvement, Rehabilitation, Maintenance, Others
V2PromoterCIdentifies the type of entity promoting or funding the project.Municipality, Others
V3RegionCIndicates the geographic region of Colombia where the project was carried out.Caribe, Pacífica, Andina, Orinoquía, Amazonía
V4Contract signing yearCIndicates the year the contract was signed.2015, 2016, 2017, 2018, 2019
Planning (6)V5Bidding valueNThe amount calculated for the selection process, not the contract value.Current Legal Monthly Minimum Wage (CLMMW)
V6Award growthNThe percentage difference between the winning bid and the bidding value for the project [64].%
V7Type of contractorCThe type or modality of the contractor who executed the contract.Consortium, Company, Individual
V8Initial contract valueNThe initial amount agreed upon in the contract, expressed in the Current Legal Monthly Minimum Wages.Current Legal Monthly Minimum Wage (CLMMW)
V9Initial durationNTime stipulated in the contract for the execution of the project.Days
V10Project intensityNThe ratio between the initial cost of a project and its duration [64].CLMMW/Days
Execution and Closure (4)V11Additional costNThe additional amount added to the initial contract value.Current Legal Monthly Minimum Wage (CLMMW)
V12Additional timeNThe time extension granted to complete the project.Days
V13Final contract valueNThe final contract amount.Current Legal Monthly Minimum Wage (CLMMW)
V14Final durationNThe final duration of the contract.Days
Table 4. Univariate analysis for dependent numerical variables.
Table 4. Univariate analysis for dependent numerical variables.
Numerical Dependent Variables
VariableMaxMinMeanSdCV
Time Deviation (%)510.00%0.00%29.00%61.00%48.00%
Cost Deviation (%)54.00%0.00%8.60%15.71%55.00%
Award Growth (%)6.55%0.00%0.24%0.71%295.00%
Project Intensity116.860.4712.1515.40126.75%
Initial Contract Value (CLMMW)20,264.7042.111516.972648.65175.00%
Initial Duration (Days)900.0010.00123.10120.8298.00%
Bidding Value (CLMMW)20,305.9242.111524.512678.14176.00%
Additional Cost (CLMMW)10,112.090.00252.591061.89420.00%
Additional Time (Days)306.000.0026.5849.68187.00%
Final Contract Value (CLMMW)30,376.7852.391769.563503.11198.00%
Final Duration (Days)900.0010.00149.70134.7590.00%
Table 5. Correlation test results between cost deviation vs. numerical variables.
Table 5. Correlation test results between cost deviation vs. numerical variables.
Project PhasesProject CharacteristicsSpearman’s Rhop-Value
PlanningInitial contract value0.211.37 × 10−3
Initial contract duration0.143.54 × 10−2
Project intensity0.185.95 × 10−3
Bidding value0.211.50 × 10−3
Award growth0.297.01 × 10−2
Execution and ClosureAdditional cost0.99<2.2 × 10−16
Additional time0.396.46 × 10−7
Final contract value0.334.25 × 10−7
Final contract duration0.326.46 × 10−7
Table 6. Correlation test results for time deviation vs. numerical variables.
Table 6. Correlation test results for time deviation vs. numerical variables.
Project PhasesProject CharacteristicsSpearman’s Rhop-Value
PlanningInitial contract value0.281.77 × 10−5
Bidding value0.281.94 × 10−5
Project intensity0.361.35 × 10−8
Execution and ClosureAdditional cost0.382.81 × 10−9
Additional time0.98<2.2 × 10−16
Final contract value0.311.92 × 10−6
Final contract duration0.451.40 × 10−12
Table 7. Kruskal–Wallis test results comparing cost deviation and categorical variables.
Table 7. Kruskal–Wallis test results comparing cost deviation and categorical variables.
Project PhasesProject CharacteristicsCategoriesMedianMean
PlanningYear20160.190.21
Others0.000.06
Project typeConstruction0.000.07
Others0.010.10
PromoterMunicipality0.000.09
Others0.000.01
Table 8. Kruskal–Wallis test results comparing time deviation and categorical variables.
Table 8. Kruskal–Wallis test results comparing time deviation and categorical variables.
Project PhasesProject CharacteristicsCategoriesMedianMean
PlanningYear20160.210.39
Others0.000.27
Contractor TypeConsortium0.100.40
Others0.000.21
Table 9. Significant project characteristics for cost and time deviation by bivariate analysis.
Table 9. Significant project characteristics for cost and time deviation by bivariate analysis.
Project PhasesProject CharacteristicsCost DeviationTime Deviation
InitiationInitial contract value
Initial contract duration
Project intensity
Bidding value
Award growth
Year
Project type
Promoter
Contractor type
PlanningAdditional cost
Additional time
Final contract value
Final contract duration
Table 10. Causes of time and cost deviations.
Table 10. Causes of time and cost deviations.
Reviewed Cause in the LiteratureCause Reported in the ProjectFrequencyTotal Frequency
Scope ChangesGreater quantities of work3088
Increase in work quantities22
Unforeseen activities20
Additional works13
Unforeseen items2
Quantity of additional works1
Weather ConditionsInclement weather2325
Inclement summer weather1
Inclement winter weather1
Changes in Original DesignsAdjustment of studies and designs49
Adjustment to original designs4
Delays in reviewing original designs1
Social ConflictsSocial conflicts68
Community interference1
Procurement difficulty due to the truckers’ strike1
Poor Planning and Project ManagementDelays in administrative tasks26
Delays in administrative procedures2
Delays in administrative processes1
Legalization of administrative procedures1
Changes in Material PricesDifficulty in negotiating materials34
Cost overruns in materials and equipment1
Increased material prices due to topography conditions2
Increased material prices due to site access conditions1
Difficulty accessing the site or materials due to poor conditions1
Shortage of LaborDifficulty hiring local labor12
Scarcity of labor1
Poor Financial ManagementEconomic adjustment of the contract11
Contractor’s Financial ProblemsEconomic adjustment of the contractor11
Table 11. Frequency of occurrence of cost and time deviation factors.
Table 11. Frequency of occurrence of cost and time deviation factors.
IDCauseFrequencyPercentage
C1Scope changes8859.40%
C2Weather conditions2516.89%
C3Changes in original designs96.08%
C4Social conflicts85.40%
C5Poor planning and project management64.05%
C6Changes in material prices42.70%
C7Difficult site access42.70%
C8Labor absenteeism21.35%
C9Poor financial management10.67%
C10Contractor’s financial problems10.67%
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Quintero, J.; Murgas, A.; Gómez-Cabrera, A.; Sánchez, O. Exploring the Relationship Between Project Characteristics and Time–Cost Deviations for Colombian Rural Roads. Infrastructures 2025, 10, 178. https://doi.org/10.3390/infrastructures10070178

AMA Style

Quintero J, Murgas A, Gómez-Cabrera A, Sánchez O. Exploring the Relationship Between Project Characteristics and Time–Cost Deviations for Colombian Rural Roads. Infrastructures. 2025; 10(7):178. https://doi.org/10.3390/infrastructures10070178

Chicago/Turabian Style

Quintero, Jose, Alexander Murgas, Adriana Gómez-Cabrera, and Omar Sánchez. 2025. "Exploring the Relationship Between Project Characteristics and Time–Cost Deviations for Colombian Rural Roads" Infrastructures 10, no. 7: 178. https://doi.org/10.3390/infrastructures10070178

APA Style

Quintero, J., Murgas, A., Gómez-Cabrera, A., & Sánchez, O. (2025). Exploring the Relationship Between Project Characteristics and Time–Cost Deviations for Colombian Rural Roads. Infrastructures, 10(7), 178. https://doi.org/10.3390/infrastructures10070178

Article Metrics

Back to TopTop