Lecture 9

Question 1

In the academic literature there are two views on how project cost overruns should be. What are these two definitions and how do they differ? What is the importance and impact of this difference?

Answer 1

In the academic literature, two main perspectives on defining project cost overruns exist:

1. Flyvberg et al. (2002) Definition: According to Flyvbjerg et al., cost overruns are measured by comparing the actual project costs to the initial estimate, which is defined as the latest available budget at the time of decision-making or project sanctioning. In this view, any deviation from this initial estimate constitutes a cost overrun. This definition emphasizes the importance of accurately estimating project costs at the outset and holding decision-makers accountable for adhering to these estimates.
2. Love and Ahiaga-Dagbui (2018) Definition: Love and Ahiaga-Dagbui challenge Flyvbjerg et al.’s definition by arguing that the initial estimate should be based on the latest budget created before project closure, rather than the budget at the time of decision-making. They contend that as projects evolve and undergo scope changes (referred to as “progressive elaboration” in project management), the original estimate becomes irrelevant, and only the final budget is a valid reference point. According to this view, cost overruns are not meaningful if measured against the initial estimate provided at the project’s outset, as this estimate may no longer reflect the project’s actual scope or circumstances.

The importance and impact of this difference lie in how it influences the perception and management of project costs. Flyvbjerg et al.’s definition may lead to a more conservative approach to budgeting and decision-making, as it emphasizes the importance of accurate initial estimates and holds decision-makers accountable for adhering to these estimates. On the other hand, Love and Ahiaga-Dagbui’s definition may result in a more flexible approach that accommodates changes in project scope and acknowledges that initial estimates may become outdated over the course of a project. The choice of definition can significantly impact how cost overruns are perceived, measured, and managed in practice, affecting project planning, execution, and evaluation.

Combination with ChatGPT

Question 2

What are Flyvbjerg’s three proposed fundamental root causes of project cost overruns? What would the statistical distribution for relative cost overrun look like for each of these root causes? Assuming Flyvbjerg’s theory to be correct, how would you test which root cause is at play in a sample of data you have at hand?

Answer 2

Flyvbjerg (2004) argues that there are three (originally four in his earlier research) fundamental root causes for cost overruns:

1. Random errors or mistakes: honest mistakes causes Gaussian noise
   * If random error or mistakes is the primary root cause, we would expect the distribution of cost overruns to be centered on zero (average = 0) and symmetric (skewness = 0)

(sett inn bilde for statistical distribution)
2. Optimism bias: see for instance Kahneman’s System 1 and 2.
* If optimism bias (or cognitive bias in general) was the root cause, we would expect the distribution of cost overruns to both have a positive mean and positive skewness in the short run, but to converge towards zero mean and skewness in the long run.

3. Strategic reporting: project managers are intentionally deceiving and manipulating for their cost estimations for either political or economic reasons
   * If strategic mistreporting is the primary root cause, we would expect the distribution of cost overruns to have a positive mean and positive skewness.

(sett inn bilde for statistical distribution)

(sett inn bilde)

To test which root cause is at play in a sample of data, you would analyze the statistical moments of the cost overruns and their temporal stability. If the distribution of cost overruns is centered around zero and symmetric, it suggests random errors or mistakes. If it has a positive mean and positive skewness, it could indicate cognitive bias or strategi misreporting. If the distribution doesn’t converge towards zero in the long run, it is more likely strategic misreporting.

Mean: measure of central tendency and represents the average value of a set of data points.
Skewness: measure the assymetry of the distribution of data arounds it’s mean. It indicates whether the data is skewed to the left or right, or if it’s symmetrical. A positive skewness value means the distribution is skewed to the right (tail is longer on the right side), while a negative skewness value means the distribution is skewed to the left (tail is longer on the left side). A skewness value of zero indicates a symmetrical distribution)

Question 3

In the context of project cost estimation, explain reverse-Darwinism – i.e. the survival of the unfittest.

Answer 3

Reverse-Darwinism, or the survival of the unfittest, in project cost estimation refers to a scenario where projects with underestimated costs are more likely to be accepted, while those with more accurate or higher estimated costs are more likely to be rejected. This happens because decions-makers often underestimate project costs, leading them to prioritize projects that seem more financially attractive but are actually more prone to cost overruns. As a result, this bias towards accepting projects with lower estimated costs leads to higher average costs for accepted projects, ultimately impacting project success and financial outcomes.