ARIMA Time Series Models for Full Truckload Transportation Prices

Jason W Miller

doi:10.3390/forecast1010009

Abstract

The trucking sector in the United States is a $700 billion plus a year industry and represents a large percentage of many firms’ logistics spend. Consequently, there is interest in accurately forecasting prices for truck transportation. This manuscript utilizes the autoregressive integrated moving average (ARIMA) methodology to develop forecasts for three time series of monthly archival trucking prices obtained from two public sources—the Bureau of Labor Statistics (BLS) and Truckstop.com. BLS data cover January 2005 through August 2018; Truckstop.com data cover January 2015 through August 2018. Different ARIMA models closely approximate the observed data, with coefficients of variation of the root mean-square deviations being 0.007, 0.040, and 0.048. Furthermore, the estimated parameters map well onto dynamics known to operate in the industry, especially for data collected by the BLS. Theoretical and practical implications of these findings are discussed.

Highlights

The trucking industry in the United States is estimated to be approximately $700 billion annually, transports nearly 71 percent of freight, and employs and estimated 2 million individuals who operate heavy duty trucks [1,2]
The ability to develop accurate pricing forecasts is challenging due to uncertainty inherent in the price of oil—which is subject to substantial swings due to geopolitical forces and technological changes—coupled with trucking services being a derived demand that fluctuates based on changes to macro-economic conditions and the seasonality of industrial production and retail sales
The autoregressive integrated moving average (ARIMA) methodology provides a useful framework for understanding the evolution of motor carrier rates because (i) the method has a substantial degree of flexibility and (ii) the theoretical meaning of ARIMA parameters map well to dynamic forces expected to affect trucking prices

Summary

Introduction

The trucking industry in the United States is estimated to be approximately $700 billion annually, transports nearly 71 percent of freight, and employs and estimated 2 million individuals who operate heavy duty trucks [1,2]. When demand for trucking services rapidly increases, prices tend to increase, as has occurred in late 2017 and throughout 2018, until more capacity comes online [6]. Given these various forces, the question that naturally arises is can accurate forecasts be generated for macro-level trucking prices? Campbell, and Mundy [8] use data from a large LTL carrier to estimate a regression model predicting revenue at the customer-lane unit of analysis using characteristics of the transported freight (e.g., total shipments, total weight of shipments, etc.). Özkaya et al [10] built a regression model to predict LTL prices using shipment, shipper, and carrier characteristics from data for approximately 485 thousand shipments over a 3-month horizon from 43 shippers and 128 carriers

Methods

Results

Discussion

Conclusion