- Resilience
- Strategy
- Transportation
- Forecasting
The United States freight spot market moves in patterns consistent with general economic fluctuations within the broader macro-economy. While the spot market does follow general trends, there is an underlying belief that pricing fluctuations occur as a result of external events. These external events, such as natural disasters, market holidays, terrorist attacks, etc., all fall into the category of “special events” due to their ability to gain a physical presence that alters socio-economic interactions. Socio-economic interactions change the market, but there has been little research into the effect that these special events have on the freight spot market. The aim of this study was to determine the underlying correlation between special events and the spot market and to discern any recognizable patterns that could help the solution providers craft effective strategies to respond to special events. A temporal-spatial multi-variable linear regression model was developed using historical spot market transactions provided by the solutions provider. The model’s results were transferred into a heat map, which was compared against a heat map of the actual values. After testing through 20 models and 3 key performance indicators, no linear correlation could be established. The highest correlation (R2) value was 0.147, which was observed in the model, Outbound Volume for Hurricane Harvey, and the lowest was 0.014, which was observed in the model, CPM for Hurricane Matthew. Linear regression has been the historical modeling technique for spot market rates, but now that the field is beginning to expand into physical and temporal regions of freight understanding, linear regression does not have the capacity to recognize the nuances that extend beyond basic economic principles. While linear regression did not provide a strong correlation between special events and the spot market in this study, future testing utilizing nonlinear techniques, such as neural networks and machine learning, is likely to produce better correlation results.