Jerry Bishop's blog post titled The Fallacy of Planning raises an interesting question. Jerry mentions cognitive bias as one source of the problem. But this bias only exists in those who fail to recognize they are biased.
But a much deeper problem exists...
The estimates prepared by cost estimators are crucially dependent on technical and programmatic assumptions over which they have little or no say . There are some gray areas; cost estimators should recognize - and provide corrections for in their estimates - some type of unrealistic program assumptions and some likely execution problems. But, without trying to fix the boundaries of these exceptions, it is clear that they are exceptions - cost estimators generally are not equipped to do engineering analyses of proposed programs or to assess the capabilities of potential contractors (McNicol, 2005, p. 19)
The examples Jerry provides of innumeracy are not the source of the problem in our domain. It may be common in IT though.
So how can IT estimimates be improved? First is to move toward what is done in our domain - have professional estimators. This is still a problem for large weapons systems. But without professional estimators or a professional grade estimating process, there is little or no help in improving the credibility of cost estimates.
Here's some background for addressing Jerry's issue - Eliminating Bias Through Reference Class Forecasting and Good Governance, Bent Flyvberg,