The purpose of this paper was to perform a detailed analysis of the challenges faced while developing and using econometric models to forecast future transportation demand. To this end, a comparative analysis of the state of practice and state of the art was undertaken on a concrete example – Viracopos Airport in Brazil. A review of relevant technical and scientific literature identified a number of approaches and each representative example was synthesized into a specification “template”. We then compared the performance of each “template” with the observed demand, through an intra-series forecast. A general finding was that econometric model specifications proved to be somewhat homogenous and simpler in nature, with results indicating a relatively small difference in fit and forecast capability across models. Even with the elimination of what is typically considered the main culprit for deviations – the forecast uncertainty of the explanatory (input) variables – the forecast is still subject to sizeable deviation. To address this issue, we proposed developing some sanity check indexes, particularly relevant for long-term forecasts. We conclude that the challenges faced at the Viracopos Airport Concession were far from econometric ones, that the success of the demand forecast and the concession itself required more than a well estimated econometric model. Finally, regarding investment obligations within the concession agreement, we strongly recommend making them conditional to meeting demand milestones, given the inherent unpredictability in forecasts.

本文的目的是对开发和使用计量经济学模型预测未来运输需求时面临的挑战进行详细分析。为此，我们以一个具体的例子（巴西的维拉科波斯机场）对实践和技术水平进行了比较分析。对相关技术和科学文献的回顾确定了多种方法，并且每个代表性示例均被合成为规范“模板”。然后，我们通过系列内预测将每个“模板”的性能与观察到的需求进行了比较。一个普遍的发现是，计量经济模型规范在本质上被证明是同质且简单的，结果表明跨模型拟合和预测能力的差异相对较小。即使消除了通常被认为是造成偏差的罪魁祸首–解释性（输入）变量的预测不确定性，该预测仍会遭受较大的偏差。为了解决这个问题，我们建议开发一些健全性检查指标，特别是与长期预测相关的指标。我们得出的结论是，维拉科波斯机场特许经营所面临的挑战远非计量经济学的挑战，需求预测的成功和特许经营本身所需要的不仅仅是一个经过良好估算的计量经济学模型。最后，考虑到预测中固有的不可预测性，对于特许协议中的投资义务，我们强烈建议使它们成为满足需求里程碑的条件。