Modeling Carrier Truckload Freight Rates in Spot Markets 5 th METRANS International Freight Conference Long Beach, CA October 8-10, 2013 C. Lindsey, H.S. Mahmassani, A. Frei, H. Alibabai, Y.W. Park, D. Klabjan, M. Reed, G. Langheim, and T. Keating
Introduction In 2012, the spot market accounted for $141.8 billion of the third-party logistics (3PL) sector. Of that, non-asset based providers accounted for $45.1 billion 1. Spot market loads are not covered by contracts. Instead, they result from a cold call from a shipper to a carrier or 3PL provider. In this study, we focus on spot market shipments brokered via non-asset based 3PLs. 1 Armstrong & Associates (2012). U.S. 3PL Market Size Estimates, http://http://www.3plogistics.com/3plmarket.htm. Accessed 10/2/2013.
Outline Problem Description Research Objectives Background Data Methodology & Analysis Shipment Understanding Profits and Losses Research Findings Research Implications
Problem Description Consider the following example: A shipper calls a 3PL provider to move a load. The 3PL broker immediately quotes the shipper a price of $100. The broker then searches for a carrier. Some agree for less than $100 (resulting in a profit), others demand more (resulting in a loss). The broker purchases capacity from the carrier that is expected to generate the most profit. The spot market is highly volatile and uncertain. Spot market shipments are fundamentally different from contract shipments.
Research Question & Objectives Research Question How do shipment characteristics affect the expected carrier truckload rates in spot markets? Research Objectives Investigate the determinants of carrier truckload rates in spot markets Determine how those determinants affect 3PL profits and losses in spot markets
Background Spot Market Procurement via Electronic Marketplaces Electronic marketplaces (ETM) offer spot market and long-term contract services Auctions have been the primary methodology to study ETMS Song & Regan (2003) Combinatorial auction for simultaneous bidding over several shipping lanes Sheffi (2004) Consideration of non-price variables in combinatorial auctions Figliozzi (2004) Sequential auctions to model an ongoing transportation spot market Motor Carrier Rates Early work concerned with the possible effects of deregulation Later work focused on less-than-truckload (LTL) operations Smith et al. (2007) Modeled net rates in order to assess discount policies Ozkaya et al. (2010) Combined quantitative and qualitative data to predict rates
Background (cont.) None of the prior studies focused on the determinants of truckload (TL) shipments None of the prior studies considered capacity procured via a search process rather than an auction This is an important segment of the 3PL industry that is largely unexplored
Data The data for this study comes from a U.S.-based 3PL provider operating in North America. Data is from the year 2011 and contains information denoting Date of the shipment Origin and destination Carrier Equipment type Price paid to the carrier Number of stops Very representative of the spot market with over 4,000 different carriers
Overall Data Process Key Components Clustering Rolling horizon
Spatial Clustering Twenty regions clustered by total shipment activity (inbound and outbound) K-means algorithm
Distance Threshold To a certain distance, carriers often charge a flat fee 1 Threshold is recovered using a regression analysis for breakpoint identification 1 Muggeo, V.M.R. 2003. Estimating regression models with unknown breakpoints. Statistics in Medicine 22 (19): 3055-3071.
Summary Statistics Variable Definition (Units) Mean 1st Quartile 3rd Quartile PPM Carrier price-per-mile ($/mi) 1.65 1.75 1.05 2.05 Distance Shipment linehaul distance (mi) 717 315 951 Variable Definition (Units) Levels Percent of Obs. Haz-Mat Shipment hazardous status Non-Hazardous > 99% Hazardous < 1% Van 75-80% Equipment Equipment type used for the move Flatbed 15-20% Refrigerated 1-2% Other < 1% 1 30-40% Calendar Quarter Quarter in which the shipment 2 25-30% occurred 3 10-20% 4 10-20% No. of Stops Number of stops required for the 2 94-97% move > 2 3-6% Lead Time Volume-to-Capacity Time between the carrier being booked and the required pick-up date of the shipment (Days) (No. of shipments on a lane / No. of carriers in a lane) during the horizon Same day 30-40% Next day 30-40% Two days 5-10% > Two days 15-25% < 5 and > 5
Shipment Analysis Methodology Linear regression model y i X i y i, X i i Shipment Price-per-mile Model parameters Shipment and Lane attributes Model errors
Shipment Analysis Results Variable Estimate t-statistic Intercept 7.405 140.6 Number of Stops 0.09631 21.49 Distance Threshold Long Dist. -3.809-55.05 Ln(Miles) -1.021-107.1 Distance Threshold Long Dist.*ln(Miles) 0.6592 53.80 HazMat Hazardous 0.5189 12.56 Equipment Flatbed 0.4044 110.4 Equipment Other 0.4708 19.91 Equipment Refrigerated 0.2278 21.56 Calendar Quarter 2 0.1173 35.80 Calendar Quarter 3 0.1062 27.59 Calendar Quarter 4 0.09742 26.50 Volume-to-Capacity High -0.04396-3.680 Lead Time High -0.1677-13.35 Lane Dummy -- -- Adj. R-Squared 0.6608 284 of the lane dummy variables are statistically significantly different from the reference lane at 5%.
Discussion of Shipment Analysis Parameter estimates of the shipment attributes (e.g., equipment type, distance, etc.) essentially capture operating costs to the carrier. Lane and market attributes (i.e., volume-to-capacity and the time of year) capture aspects of carrier rates not directly related to carrier costs. These are likely the variables that largely determine carrier profit margins, the aspect of cost to 3PL providers that could be negotiable.
Understanding Profits and Losses Methodology Linear regression model y i X i y i, X i i Shipment Net profit Model parameters Shipment and Lane attributes Model errors
Understanding Profits and Losses Results Variable Volatility Volatility Threshold Volume Definition Standard deviation of PPM on a given lane within the horizon Breakpoint at which there are decreasing returns to volatility Number of shipments on a given lane within the horizon Variable Estimate t-statistic Intercept 0.7238 42.20 Volatility 0.02929 3.916 Volatility* Volatility Threshold -0.03060-3.998 Volume -0.0006204-15.87 Miles -0.0003726-39.73 HazMat Hazardous 0.2111 5.039 Equipment Flatbed 0.02689 7.265 Equipment Other 0.003574 0.1490 Equipment Refrigerated -0.02504-2.341 Calendar Quarter 2-0.02721-8.079 Calendar Quarter 3 0.007340 1.837 Calendar Quarter 4 0.01598 4.230 Lane -- -- Adj. R-Squared 0.2162
Understanding Profits and Losses Discussion Increased exposure ( Volume ) translates to an increased likelihood of suffering a loss with greater opportunity for an unprofitable transaction, losses are realized more often and with greater magnitude. Volatility is associated with profitable transactions for values less than the threshold, and losses for values above the threshold. To a point, 3PL providers are able to exploit market uncertainty to their benefit until falling victim to it themselves.
Research Findings & Implications The information derived from this study is useful to 3PL providers operating in spot markets Yields greater insight into spot market carrier rates beyond lane averages on online databases Provides a methodology for processing spot market data Potentially diagnose causes of unprofitable transactions
Final Thoughts In the future, this work could be improved by Pooling data across several freight forwarders in order to gain a more complete perspective of the market. Testing alternative clustering algorithms Considering that political boundaries may have a greater influence on carrier rates than assumed
Questions? Contact Information The Transportation Center Northwestern University www.transportation.northwestern.edu Christopher Lindsey clindsey@u.northwestern.edu