The impact of railroad mergers on grain transportation
markets: a Kansas case study
Joon Je Park
a,1, Michael W. Babcock
b,*, Kenneth Lemke
c,2 aDepartment of Agricultural Economics, North Dakota State University, Fargo, ND 58105, USA b
Department of Economics, Kansas State University, Manhattan, KS 66506, USA c
Data Transmission Network Corporation, 9110 West Dodge Road, Suite 200, Omaha, NE 68114, USA
Received 14 July 1998; received in revised form 28 March 1999; accepted 8 June 1999
Abstract
While there have been many studies of the impact of railroad deregulation on agricultural transportation markets there have been very few that address the impact of railroad mergers on rail grain prices and the distribution of eciency gains. The purpose of this paper is to add to the sparse literature regarding the eect of railroad mergers on agricultural transportation markets. Given the ever declining number of Class I railroads, this research is very timely.
The speci®c objectives of the research are as follows: (1) Analyze the impact of the Burlington Northern (BN)±Santa Fe (SF) merger on the ability of the BNSF to increase prices on movements of Kansas wheat to Houston, Texas. (2) Analyze the impact of the Union Paci®c (UP)±Southern Paci®c (SP) merger on the ability of the UPSP to increase prices on movements of Kansas wheat to Houston, Texas. (3) Analyze changes in Kansas wheat logistics system costs as a result of the BN±SF and UP±SP mergers.
Two models are developed to achieve the objectives of the study. A network model of the wheat logistics system is used to identify the least cost transportation routes from the Kansas study area to the market at Houston, Texas. A pro®t improvement algorithm is developed to measure the amount by which railroads can raise their prices above variable cost.
The BNSF and UPSP achieve only minor increases in market power (measured by the ratio of revenue to variable cost) because the merged railroads have only slight advantages in cost relative to other railroads that serve the same areas as the merged railroads. Wheat shippers bene®t from merger-induced reductions in transportation and handling costs. Shippers are likely to capture a signi®cant share of these cost re-ductions since intrarailroad competition is present after the mergers. Transport cost rere-ductions accompany
*Corresponding author. Tel.: +785-532-4571; fax: +785-532-6919. E-mail address:[email protected] (Michael W. Babcock)
1
Tel.: +701-231-7498; fax: +701-231-7400 2
Tel.: +402-390-2328; fax: +402-255-3667
1366-5545/99/$ - see front matter Ó 1999 Elsevier Science Ltd. All rights reserved.
mergers due to more direct routing of wheat shipments and the assumption that the merged railroad op-erates at the costs of the lower cost partner. Ó 1999 Elsevier Science Ltd. All rights reserved.
1. Introduction
The Staggers Rail Act of 1980 greatly reduced regulation of the railroad industry. The act recognized the high degree of intermodal competition in most transportation markets. Also in 1980 there were 39 Class I railroads, indicating potential intramodal competition. Thus compe-tition rather than regulation would determine rail prices.
The Staggers Rail Act also contained provisions designed to expedite railroad mergers. For example, merger decisions by the ICC had to be made within 30 months of the date of application; this was later reduced to 15 months by the 1995 ICC Termination Act. Although the impact of deregulation is dicult to measure, the number of Class I railroads has declined from 39 in 1980 to nine in 1998.
Proponents of mergers cite several potential advantages. Some of these would bene®t shippers while others would assist merging carriers. In end-to-end mergers, shippers have direct access to more markets on the merged system. Reduced need for interlining also results in faster delivery time and less administrative costs for shippers. In parallel mergers the elimination of duplicate facilities reduces the costs of the combined system. Parallel mergers will reduce rail costs and if there is strong competition, this will turn rail cost reductions into reduced prices.
Market power was the major reason for regulating rail prices. The recent rapid consolidation of the rail industry concerns many observers including many who initially supported the Staggers Act. Critics of rail mergers argue that the potential cost savings are overstated and that the primary eect will be an increase in rail prices. They contend that mergers will not produce re-source savings but they merely transfer income from shippers to railroads.
There have been many studies of the impact of railroad deregulation on agricultural trans-portation markets. These include Adam and Anderson (1985), Babcock (1981, 1984), Babcock et al. (1985), Casavant (1985), Chow (1984), Fuller et al. (1983), German and Babcock (1982), Klindworth et al. (1985); MacDonald (1987, 1989), US Department of Agriculture (1982, 1984), Sarwar and Anderson (1986), Fuller et al. (1987), Sorenson (1984), Wilson (1985), Hauser (1986), Thompson et al. (1990), Kwon et al. (1994) and Schmitz and Fuller (1995). However there have been comparatively few studies of the eects of rail mergers on agricultural transportation markets. The few that exist have examined railroad eciency gains, market share changes, and the eect of merger con®guration on eciency gains (i.e. end-to-end vs. parallel mergers). Only Le-mke and Babcock (1987) directly address the impact of railroad mergers on rail grain prices and the distribution of eciency gains. The purpose of this paper is to add to the sparse literature regarding the eect of railroad mergers on agricultural transportation markets. Given the ever declining number of Class I railroads, this research is very timely.
The objectives of the paper are as follows:
1. Analyze the impact of the Burlington Northern (BN)±Santa Fe (SF) merger on the ability of the BNSF to increase prices on movements of Kansas wheat to Houston, Texas.
3. Analyze changes in Kansas wheat logistics system costs as a result of the BN±SF and UP±SP mergers.
2. The study area
The grain origin area of this study includes all of 57 counties and parts of 15 other counties in central and western Kansas that comprises about two-thirds of the Kansas land area (Fig. 1). This
area is divided into 342 production origins (12´12 square mile areas).
In 1996 the area produced about 217 million bushels of wheat which was 85% of the stateÕs
production.3The area also produced about 77% of total Kansas sorghum production and 81% of
the stateÕs corn production.4
The area is served by four Class I railroads including Burlington Northern (126 miles), Union Paci®c (852 miles), Santa Fe (658 miles), and Southern Paci®c (259 miles). In addition there are ®ve short lines serving the area (not shown in Fig. 1) including Kyle Railroad (632 miles), Kansas Southwestern (302 miles), Central Kansas Railroad (882 miles), Cimmaron Valley (182 miles), and the Garden City Western (45 miles). Thus the total miles of rail line in the study area is 3938. These railroads serve study area elevators with a total storage capacity of 670 million bushels.
The study focuses on wheat because rail transportation is the dominant transportation mode for Kansas export wheat. This is not the case for corn or sorghum. For the 1992±1993 crop, small country elevators moved 52% of their wheat shipments by rail. The corresponding percentages for
large country elevators and terminal elevators were 69% and 99%, respectively.5
Most of the wheat grown on Kansas farms travels though a system of country elevators and inland terminal elevators to domestic ¯our mills or export terminals at coastal ports. Port ter-minals on the Gulf of Mexico have been the major destination for Kansas export wheat. During the 1992±1993 crop year, 62% of all Kansas wheat shipments were to Gulf of Mexico ports.
The wheat logistics system in Kansas is summarized in Fig. 2. Trucks are used to ship wheat from farms to country elevators, inland terminal elevators, or subterminal elevators. Country elevators ship wheat to inland terminals or domestic ¯our mills by truck or rail. Inland terminal elevators and subterminal elevators ship wheat by rail and by truck to other inland terminals or to ®nal markets. Inland terminal elevators have functioned as intermediate collection, storage, and routing points. Subterminals are designed for rapid loading of unit trains of wheat.
In the study area, inland terminal elevators are located at Salina, Wichita, and Hutchinson, Kansas. Subterminals are located at Colby, Dodge City, Wakeeney, Liberal, and Ogallah, Kansas and Enid, Oklahoma.
Houston, Texas was selected as the only market in order to make the results of the study more realistic. The study uses a transportation cost minimizing algorithm to obtain least cost wheat movements. The model minimizes total transportation and handling costs for the movement of wheat from production origins to Houston, Texas. If other destinations had been included as
3
Computed from data in Kansas Department of Agriculture (1997). 4
Computed from data in Kansas Department of Agriculture (1997). 5
markets, the least cost solution would indicate that some production origins do not ship wheat to Houston. This is undesirable for three reasons. First, an objective of the study is to determine the ability of railroads to increase prices on wheat shipments to Houston. To do this it is necessary to
®nd the least cost route fromeachstudy area production origin to Houston. Second, rail service to
Houston is available from every elevator site in the study area, so railroads have established prices for these movements. Third, in the 1992±1993 crop year over 60% of Kansas wheat shipments
moved to Gulf of Mexico ports for export, and signi®cant amounts moved from each production origin.
Exclusion of non-Gulf port areas will not prevent achieving the objectives of the study. This is because those other ports are served by other wheat supply areas which have a locational ad-vantage. Thus the Paci®c Northwest ports are served by Washington, Montana, and the Dakotas. California ports are served by Utah, California, and Arizona. The western Great Lakes ports are served by Montana, the Dakotas, and Minnesota.
3. Models and procedures
Two models are required to achieve the objectives of the study. A network model is required to identify the least cost transportation routes from the study area to Houston. The network model is a variation of the linear programming methodology. A pro®t improvement algorithm is needed to measure the amount by which railroads can raise their prices above variable cost in the face of changing market conditions.
3.1. The network model
The network model includes 342 production origins (12 miles by 12 miles in area), 280 country elevators, 6 subterminals, 3 inland terminals, and one destination (Houston). The initial movement of wheat is from a production origin to a country elevator, terminal elevator or a subterminal. A production origin may ship wheat to any country elevator that is within a 50 mile
radius; it may ship to a subterminal if it is within a 70 mile radius. A production origin can ship to any terminal. Because one of the objectives of this study is to examine the ability of railroads to increase their market power and pro®ts, production origins must be allowed to deliver Kansas wheat farther than historical average mileages in order to identify transportation costs of
al-ternative routes.6
In the network model, wheat moves through the logistics system to the ®nal demand point. After the wheat is assembled at country elevators or subterminals, the following types of shipment patterns may occur (Fig. 3):
1. Multicar (15 car) rail shipments from country elevators to Houston, Texas.
2. Multicar (15 car) rail shipments from country elevators to inland terminals or subterminals. This is followed by a 100 car unit train movement from inland terminals or subterminals to Houston, Texas.
3. Commercial truck movement from country elevators to inland terminals or subterminals. The subsequent movement is a unit train shipment of 100 cars from inland terminals or subtermi-nals to Houston, Texas.
4. Commercial truck movement from country elevators served by a given railroad to another country elevator served by a competing railroad (s). This is followed by multicar (15 car) ship-ments to inland terminals or subterminals. The next movement is 100 car unit train movement from inland terminals or subterminals to Houston, Texas.
6
5. Commercial truck movements from country elevators served by a given railroad to an inland terminal or a subterminal served by a competing railroad (s). The subsequent shipment is unit train shipment of 100 cars from inland terminals or subterminals to Houston, Texas.
Intramodal competition between Class I railroads is simulated by not allowing Class I railroads with routes to Houston, Texas to transfer wheat trac to another Class I railroad. Class III railroads that do not have direct routes to Gulf ports are allowed to transfer wheat shipments to railroads that are not competitors for the same production origins. For example, the Kyle in-terlines with the Santa Fe before the BN±SF merger and with the BNSF after the merger but not with the Union Paci®c. This is because the Kyle and the Union Paci®c both serve the northern portion of the study area while Santa Fe does not.
The objective function of the network model is to minimize total logistics system costs, which
include grain transportation and handling costs, subject to the following constraints:7
(1) The ¯ow of wheat into each transshipment point exactly equals the ¯ow out of the trans-shipment point.
(2) The total amount of wheat supplied exactly equals the total amount of wheat demanded. (3) All grain transportation and handling cost coecients are greater than or equal to zero; all endogenous variables are greater than or equal to zero.
A transportation cost minimizing algorithm is used to solve the network model for the least cost movements of wheat. This algorithm requires that the following capacity constraints be added to the model:
(4) Storage at each facility must be equal to the storage capacity (total storage capacity of all the elevators at the location is used).
(5) The quantity of wheat shipped by each mode between each location must be less than or equal to the total storage capacity at the origin location.
(6) No grain stocks will remain at the farm or at transshipment points at the end of the crop year.
The cost minimizing algorithm is used to solve the network model.8
3.2. The pro®t improvement algorithm
The pro®t improvement algorithm is used to evaluate each railroadÕs ability to raise prices
above variable costs. The algorithm simulates a range of prices from variable cost to a level that diverts all trac to rival railroads or other transport modes. The algorithm then identi®es the set
of constrained Nash Equilibrium prices: a set of prices that maximizes the railroadÕs net revenues
subject to the constraint that the prices of all rival railroads are set equal to their respective variable costs.
7
For a detailed discussion of the mathematics of the network model, see Park (1998), pp. 80±83. 8
The pro®t improvement algorithm starts with all railroad rates set equal to variable costs and systematically generates changes in rates for a given railroad that will increase its excess of rev-enues over variable costs. The algorithm uses the solution from the network model to calculate shadow prices for alternative routes from production origins to ®nal destination. Shadow prices represent the increase in total transportation costs if one unit of grain is shipped over a route that
is not part of the least cost solution. Shadow prices indicate a railroadÕs cost advantage over its
competitors on a speci®c route. In the pro®t improvement algorithm, shadow prices are used to
determine the amount a railroadÕs prices may be increased without loss of trac on each of its
routes. The algorithm also evaluates the impact of raising prices to levels where some trac is diverted to other carriers.
In this study, railroad rates are initially set equal to variable costs. The rates of competing railroads are also held at variable cost. In Phase I, the algorithm increases the prices of the railroad being analyzed by an amount equal to the cost disadvantage of the lowest cost alternative route. In Phase II, the algorithm increases prices above those in Phase I to levels that divert some wheat shipments to competitor carriers. Hence, in Phase II, the total volume of wheat shipped by
the railroad being analyzed is decreased relative to the railroadÕs volume in Phase I. The price
increases are terminated when the algorithm reports no additional contribution to net revenues.
3.3. Procedures
In this study, a procedure for evaluating the potential eects of Class I railroad mergers is developed. The ®rst step in this procedure is to identify geographic areas where each railroad has a cost advantage in the movement of Kansas wheat. This is done by ®rst ®nding the least cost solution to the network model and then determining the geographic area served by individual railroads.
The second step is to determine the cost disadvantage of the least cost carrierÕs best competitor
for each wheat origin and the cost disadvantage of other routes the least cost carrier could use to provide service to each wheat origin. The least cost solution to the network model is used to determine these values.
The third step is to derive estimates of the least cost carrierÕs ability to raise prices when
competitor prices are held constant at variable cost. By assuming that rival railroads set price at
variable cost, the simulation imposes themaximum constrainton the ability of study area railroads
to increase their Kansas export wheat transportation prices.
The fourth step is to simulate changes in competition. Mergers are simulated by removing the restriction that carriers cannot interchange trac and by eliminating interchange charges between merging carriers.
The ®fth step is to estimate the carrierÕs ability to raise prices after the mergers. The procedure
used to derive these estimates is the same as in steps one through three outlined above for the pre-merger railroads.
Finally, the estimates of each railroadÕs market power (ability to raise prices) after the mergers
is compared with its market power before the mergers. In this study, market power is de®ned as the ratio of rail prices to rail variable costs. Market power is reported for each railroad aected by mergers. The market power value reported for a railroad is calculated by dividing the estimated
measure of market power was selected because it was the ratio used by the former Interstate Commerce Commission to test for rail market dominance.
Two estimates of market power and net revenue (railroad revenue minus variable cost) were calculated. The ®rst estimates are the solutions to the ®rst iteration of Phase I of the pro®t improvement algorithm. These are the market power and net revenue estimates of the railroads if they restrict themselves to the movements of the least cost solution. The second set of esti-mates are the solutions to Phase II of the pro®t improvement algorithm which allows railroads to deviate from least cost routes if net revenue can be increased by doing so. Since the pattern of results is the same for both phases, in the interest of brevity only the Phase II results are re-ported.
4. The data
The network model requires the speci®cation of several predetermined structural variables including the following:
(a) the number of storage periods within the crop year;
(b) the size and number of production origins in the study area; (c) sites and storage capacities of country elevators in the study area; (d) sites of inland terminals; and,
(e) sites of export terminals.
Multiple crop year time periods are not appropriate for this study. The objectives of the paper require identi®cation of geographic areas where individual railroads have a cost advantage. Ef-fective storage capacity constraints at production origins and multiple time periods would force the network model to select other than least cost routes for some wheat shipments. Therefore, a model with multiple time periods would not provide accurate information about the potential
market areas of individual railroads or the ability of a railroad to restrict a competitor railroadÕs
prices.
Each production origin is represented by a 12 by 12 square mile area resulting in 342 farm production origins in the study area.
A total of 280 locations (including 13 country elevators in Nebraska) were selected for country elevator sites based on the elevator size and the presence of rail service. A total of 46 locations were not included in the network model because they have no rail service. The storage capacities for country elevators are obtained from Kansas Grain and Feed Association (1994).
Five elevator locations (Colby, Dodge City, Liberal, Ogallah, and Wakeeney, Kansas) in the study area and one elevator location (Enid, Oklahoma) out-of-state were selected from country elevator locations in the study area to be subterminal sites. A country elevator is considered a subterminal site if an elevator facility at that location has railroad service and has sucient storage and loading capacity for 100 car shipments.
Houston, Texas is chosen as the export terminal and is used to measure the shipping distances from the study area.
The network model requires exogenous estimation of the quantity of wheat supplied for export from each of the 342 production origins. The average Kansas wheat production of 1994, 1995 and 1996 is assumed to be the total wheat supply. During the assumed time frame, the study area produced an average 283.93 million bushels of wheat which was 87.4% of the total Kansas wheat production. The three year average production of each study area county is divided by the area of the county to obtain the average production for each square mile. The wheat supply for export of each production origin is then estimated by multiplying the average production of each square mile by the number of square miles in the production origin. The wheat production data is from
Kansas Farm Facts(Kansas Department of Agriculture, 1995, 1996). By assumption, the quantity
of wheat demanded at Houston is equal to the total three year average production (284 million bushels) of the study area.
The network model requires transportation cost estimates for farm trucks, commercial trucks, and railroads. Costs are used rather than prices because the objectives of the paper require
es-timates of the least cost carrierÕs cost advantage relative to its competitors. Also cost estimates are
needed to calculate revenue/variable cost ratios, our measure of market power.
In this study, it is assumed that movements from production origins are by custom wheat harvester trucks. The prices charged for these wheat shipments are from Kansas Department of Agriculture (1994). These trucks have a high percentage of variable costs so there is little dier-ence between prices and costs.
Commercial truck costs are assumed to have a cost advantage relative to railroads for short hauls. This is because truck costs increase in direct proportion to distance whereas rail costs do not due to the high percent of ®xed costs in railroad cost structures. Thus after the short line railroad cost function is estimated, the commercial truck cost function is assumed to have a cost advantage relative to the short line for 0±50 miles, but higher costs than the short line for dis-tances that are greater than 50 miles. The estimated commercial trucking cost function used in the
model is based on Berwick (1997) and is as follows:9
Dollars/1000 Bushels9.8 + 0.7376 (one way mileage).
The mileage between country elevators, subterminals, and terminals in the study area is esti-mated by aggregating the highway map distances between origins and destinations.
Estimates of Class I railroad variable costs are from the ICC Railroad Costing Program (ICC, 1991). The 1994 costs of BN, SF, UP, and SP are used for wheat shipments from the study area to Houston. When these railroad merge, the combined railroad is assumed to operate at the costs of the lowest cost partner. This assumption has a parallel in the merger simulation literature (We-rden and Froeb, 1996, p. 72 and We(We-rden and Froeb, 1994, p. 415). For example, since BN has lower costs than SF, the BNSF is assumed to operate with the BN cost structure. This assumption
9
seems reasonable since there is reduced economic incentive for a merger if the merging ®rms are unable to reduce their costs.
The results of the analysis are not sensitive to the costs that are assumed for merged railroads. According to the data in Table 1, the revenue/variable cost ratios, corresponding to the as-sumption that the merged railroad experiences no cost reduction, are virtually identical to the ratios obtained by assuming the merged railroad operates at the costs of the lowest cost partner. This indicates that most of the post-merger cost reductions observed in the empirical section of the paper are due to the more direct routing enabled by the mergers. Cost reductions due to econ-omies of scale may also occur in merged railroads. Unfortunately the only empirical rail costing data (ICC, 1991) available is unable to measure economies of scale. Thus the analysis in this paper is only able to measure post-merger cost reductions due to more direct routing and operation of the merged railroad at the costs of the most ecient partner.
Short lines have lower costs than Class I railroads. The three major sources of cost savings are labor, equipment, and maintenance of way. Labor costs are lower due to more ¯exible work rules, smaller crews, and lower wages and bene®ts. Short lines formed since 1970 operate with an
av-erage of 0.54 employees per mile of track compared with 1.8 employees for Class I railroads.10
Therefore, short line railroads are able to operate rail lines at a lower cost than that of Class I railroads.
Since the ICC costing program only measures Class I rail costs and there is no other cost function for short lines, some assumption must be made regarding the measurement of short line Table 1
Eect of post-merger cost assumptions on post-merger railroad revenue/variable cost ratios
Railroad Revenue/variable cost ± railroads retain their pre-merger costs
Revenue/variable cost ± merged railroad assumes the costs of the low cost partner
railroad costs. Among Class I railroads, Illinois Central (IC) has the lowest cost structure. Thus the cost function of IC is regarded as the short line railroad cost function.
Grain handling costs were supplied by Mack N. Leath of the Economic Research Service, US Department of Agriculture.
5. Empirical results
Before discussing the impact of the BN±SF and UP±SP mergers on railroad costs and prices it is useful to discuss the nature of the results generated by the network model and the pro®t im-provement algorithm. While the estimates of rail variable costs derived in this study are estimates of actual costs, the values derived for rail revenues and prices are not estimates of actual revenues or prices. These values underestimate actual rail prices and revenues because they are derived under the stylized assumption that competitor railroads set prices that are equal to variable costs. While these estimated prices do not provide estimates of actual railroad prices, they do provide an indication of expected trends in price/variable cost ratios (measures of market power) due to railroad mergers.
There are alternative methodologies for evaluating the impact of railroad mergers on railroad grain prices. For example, regression analysis could be performed with actual railroad prices as the dependent variable. The explanatory variables could include various railroad demand and cost factors as well as merger eects (perhaps proxied by a dummy variable). The diculty with this approach is that actual rail price data are not available by year and by individual railroad. To empirically estimate the regression model would require actual railroad price data for many years prior to and after the railroad merger. In addition the actual railroad price data would have to be
railroad speci®c to examine the eects of mergers. Railroad revenue is published in Carload
Waybill Statisticscompiled by the Association of American Railroads, however the data are not
available by railroad. Also, Wolfe and Linde (1997) raises serious questions concerning the use of Waybill data for railroad price studies.
The objective of this study is not to measure actual railroad price changes resulting from
mergers but rather changes in the ability of merged railroads to raise their prices. The
method-ology used in the study was selected because it systematically measures the ability of merged railroads to raise their prices. Nevertheless we indirectly measured actual railroad prices both before and after the mergers to determine if actual railroad price changes are consistent with the empirical results of the analysis.
The study area is divided into four subregions to facilitate handling of the large computational requirements of the network model (Fig. 1). Results of these four subregions are aggregated to evaluate results for the whole study area logistics system.
5.1. Impact of the BN±SF merger in subregions 1 and 2
The UP obtains no bene®t from the merger with SP in subregion 1. This is because UP has more direct routes to Houston than SP, and UP has lower costs than SP.
subregion 1. This occurs because the cost of shipping wheat to the Gulf of Mexico on the BN is greatly reduced due to more direct routing. Also it is assumed that after the merger the combined railroad assumes the costs of the lowest cost partner. Thus since BN costs are lower than SF costs, BNSF is assumed to operate at BN cost. Therefore, most of the increase in market power and net revenue occurs because the combined BNSF railroad is assumed to be a more ecient carrier.
Table 2 contains estimated net revenue for the railroads serving subregion 1 before and after the BN±SF merger after least cost routes have been recalibrated by the pro®t improvement algorithm.
As the table indicates, BN net revenue (revenueÿvariable costs) increased from $101.80 per 1000
bushels before the merger to $112.40 after the merger. In contrast, Kyle net revenue plunged from $121.00 per 1000 bushels before the merger to only $15.30 after the merger. The corresponding ®gures for UP are $57.20 (pre-merger) and $41.50 (post-merger).
The market power (revenue/variable cost) of BN increases from 1.04 (pre-merger) to 1.057 (post-merger), while the corresponding ®gures for Kyle are 1.053 and 1.048. Likewise the market power of UP falls from 1.042 to 1.038. These ratios are well below that which is generally accepted as the
minimum price needed to cover total rail costs.11 These low ratios are attributable to the
nor-malization assumption that both UP and Kyle are setting a price equal to variable cost. Therefore, the combined BNSF still faces a high degree of competition from the Kyle and UP railroads.
While the subregion 1 ratios of rail revenue to variable costs show changes in market power, they do not provide a complete picture of the impact of the BNSF merger on the aected rail-roads. The impact on trac volume is also important.
As a result of the BN±SF merger, the revenue/variable cost ratio of the Kyle falls from 1.053 to 1.048. This appears to be a relatively insigni®cant amount. However, this measure of market power does not re¯ect the fact that the Kyle is no longer the least cost carrier for most of the wheat origins where it had a pre-merger cost advantage. The total volume of wheat shipped by the Kyle in subregions 1 and 2 falls from 39 040 thousand bushels to 29 176 thousand bushels, a decline of 25%. Railroads have a high percentage of ®xed assets, and a signi®cant reduction in trac volume would force a railroad to either raise prices, due to higher cost per unit shipped, or accept reduced pro®ts (or increased losses) until the size of the railroad could be adjusted to new trac levels. Thus, the potential reduction in trac volume is far more important to the management of the Kyle than the Table 2
Estimated net revenue and revenue/variable cost ratios for aected railroads before and after BN±SF merger when least cost routes are recalibrated (subregion 1)
change in average revenue/variable cost ratio. It is very likely that the Kyle would strenuously resist any major loss of trac by improving service and/or reducing price.
The results of the model show that after the merger the BNSF should dominate the movement
of wheat from subregion 1. However, the BNSFÕs cost advantage relative to Kyle and UP is
relatively small and the Kyle cannot aord to lose a signi®cant volume of its trac. Therefore, there should be signi®cant cost reductions and a very high level of competition among the carriers after the merger. The ®nancial position of the Kyle could be seriously threatened.
The BNSF merger should bene®t most shippers. The cost advantage of the combined BNSF will allow it to compete for more trac. This increased competition should drive down prices for most wheat shippers in the short run. The long run eects are not as clear. The combined BNSF has a cost advantage for most of the wheat shipments from the north portion of subregion 1. This advantage may enable it to drive the Kyle from this market. If this occurs, shippers could face increased prices in the long run.
Table 3 contains total transportation and handling costs plus estimated pro®ts from Phase II of the pro®t improvement algorithm for aected origins in subregion 1. As the data in the table indicate, these values decline after the merger for all aected production origins in subregion 1, with an average decrease of 6.25%. To the extent that lower costs and limitations on the estimated ability of railroads to raise prices translate into reductions in actual rail prices, wheat shippers would bene®t from the BN±SF merger.
Table 4 contains estimated net revenue for the railroads serving subregion 2 before and after the BN±SF merger after least cost routes have been recalibrated by the pro®t improvement algorithm. The merger has virtually no impact on the pro®ts of Central Kansas Railway (CKR) and the UP.
Table 3
Eect of BN±SF merger on transportation and handling costs plus estimated pro®ts when railroads raise prices and recalibrate least cost routes (subregion 1)
The net revenue of BN increases somewhat while Kyle net revenue decreases from $101.30 per 1000 bushels (pre-merger) to $85.40 (post-merger). The same pattern occurs in the railroad rev-enue/variable cost ratios. There is virtually no change in the market power (revrev-enue/variable cost ratio) of CKR and UP, but market power of Kyle declines somewhat from 1.05 (pre-merger) to 1.042 merger). The market power of BN increases from 1.038 (pre-merger) to 1.058 (post-merger). Thus there is a redistribution of pro®ts and market power in subregion 2 from the Kyle to BN. This is because the former BN is able to reduce its costs after the merger due to more direct routing. Also the former SF in subregion 2 assumes the lower costs of the BN after the merger. However the BNSF cost advantage relative to the other railroads serving subregion 2 is small and the combined railroad still faces a high degree of intramodal competition.
5.2. Impact of the BN±SF and UP±SP mergers in subregions 3 and 4
Since there are no former BN lines in these subregions the only eect of the BN±SF merger is that the SF assumes the lower cost structure of the BN.
The merger between the UP and the SP allows the SP to use more direct UP routes to the Gulf of Mexico for the wheat it can obtain on its line from Liberal to Hutchinson, Kansas, signi®cantly
lowering the SPÕs costs. Also it is assumed that the combined railroad assumes the lower costs of
the UP.
Table 5 contains subregion 3 estimated net revenue (revenue±variable costs) and revenue/ variable cost ratios for aected railroads before and after the BN±SF and UP±SP mergers when least cost routes are recalibrated by the pro®t improvement algorithm. As the data in the table
Table 5
Estimated net revenue and revenue/variable cost ratios for aected railroads before and after BN±SF and UP±SP mergers when least cost routes are recalibrated (subregion 3)
Railroad Pre-merger Post-merger
indicate, UPSP net revenue jumps from $2.80 per 1000 bushels before the two mergers to $139 after the mergers. Net revenue of BNSF increases somewhat from $45 per 1000 bushels before the mergers to $49.70 after the mergers. This result is due to the fact that the former SF is assumed to operate at the lower costs of the BN after the BN±SF merger. The reduced costs allow the former SF to divert trac from Cimmaron Valley (CV) railroad. The net revenue of CV plunges from $107.90 (pre-merger) to only $14.20 (post-merger). Thus the mergers could threaten the long term viability of CV.
Table 5 also reveals that the subregion 3 market power (revenue/variable cost ratio) of UPSP increases from 1.013 (pre-mergers) to 1.061 (post-mergers). The market power of BNSF rises from 1.036 to 1.050 while that of the CV falls from 1.052 to 1.041. Thus while the market power ratio of the UPSP increases after the mergers, it is still only slightly greater than one. This is because of the normalization assumption that the BNSF and CV set their prices at variable cost.
Table 6 contains total transportation and handling costs plus estimated pro®ts from Phase II of the pro®t improvement algorithm for aected origins in subregion 3. As the data in the table indicate, these values decline after the mergers for all aected production origins in subregion 3, with an average decrease of nearly 25%. To the extent that lower costs and limitations on the estimated ability of railroads to raise prices translate into reductions in actual prices, wheat shippers would bene®t from the BN±SF and UP±SP mergers.
The reduction in transportation and handling costs is greater in subregion 3 than subregion 1 because there are more sources of saving in these costs. In subregion 3, the former SF assumes the lower costs of the former BN in the combined railroad while the former SP assumes the lower costs of the former UP. Also the SP routes to Houston are very circuitous so its merger with UP resulted in large cost reductions due to more direct routing.
Table 7 contains estimated net revenue for the railroads serving subregion 4 before and after the BN±SF and UP±SP mergers after least cost routes have been recalibrated by the pro®t im-provement algorithm. The mergers have almost no impact on the net revenue of CKR and Kansas
Table 6
Eect of UP±SP and BN±SF mergers on transportation and handling costs plus estimated pro®ts when railroads raise prices and recalibrate least cost routes (subregion 3)
Southwestern (KSW) while the net revenue of BNSF increases somewhat and the net revenue of UPSP increases substantially from $54.30 per 1000 bushels (pre-mergers) to $89.20 (post-merg-ers). Net revenue of BNSF increases because it is assumed that the former SF operates at the lower costs of the BN after the merger. Net revenue of UPSP increases as a result of the SP obtaining the direct routes of the UP to Houston and by operating at the lower costs of the UP. The BN±SF and UP±SP mergers produce virtually no change in the market power of the four railroads serving subregion 4. This is likely due to strong intermodal competition from com-mercial trucks which have a cost advantage for short hauls to terminal elevators at Wichita and Hutchinson, Kansas as well as Enid, Oklahoma.
6. Estimating actual railroad prices before and after mergers
The foregoing investigates potential impacts on pricing by merged railroads on the assumption that competitor railroads price at variable cost. A more direct approach is to examine actual rail pricing before and after a merger. But reliable actual railroad price data do not exist. However according to Fuller et al. (1987) grain prices in an export grain corridor re¯ect the port price and the transportation and marketing charges necessary to market the grain to the respective port. Thus the dierence between the Houston wheat price and the Kansas farm level wheat price is equal to transportation and marketing charges for shipping wheat from Kansas to Houston. This is called the geographic price spread which is computed as follows:
Price SpreadHouston Wheat PriceÿKansas Wheat Price
Transportation CostsMarketing Costs:
This assumes that the grain handling industry is suciently competitive so that any change in transportation price will be re¯ected in the geographic price spread. There is empirical support for this assumption. Babcock et al. (1985) found that changes in export rail wheat transport prices from selected Kansas origins to the Gulf of Mexico are nearly identical to changes in geographic price spreads between Kansas and the Gulf of Mexico. Klindworth et al. (1985) found the transportation price to be the largest component of the Kansas-Gulf of Mexico price spread; for 14 Kansas grain elevators the export rail wheat price averaged 84±92% of the geographic price spread. Thus we can be reasonably sure that the Kansas-Houston geographic price spread is Table 7
Estimated net pro®ts and revenue/variable cost ratios for aected railroads before and after BN±SF and UP±SP mergers when least cost routes are recalibrated (subregion 4)
approximately the same as the actual rail wheat transportation price from Kansas origins to Houston for export.
Table 8 contains geographic price spreads for 52 grain elevator locations in the study area. The BN±SF merger was completed in September 1995 while the UP±SP merger was consummated in July 1996. The pre-merger price spreads in Table 8 are the average price spreads (measured in cents per bushel) for the second quarter of 1994 while the post-merger price spreads are the av-erage price spreads for the second quarter of 1998.
An examination of the table indicates that only eight of the 52 price spreads increased after the
mergers and only two of the eight increased by more than 5%. The other 44 price spreads
de-creasedafter the merger with 17 of the 44 price spreads decreasing by more than 10%. Thus the
analysis indicates that actual railroad wheat transport prices in the study area declined following the mergers which is consistent with the conclusion of the network model analysis that mergers did not increase the ability of merged railroads to raise their prices. Stated another way, declining post-merger railroad prices are not consistent with the hypothesis that mergers increase railroad market power.
The results of this study are consistent with those of Lemke and Babcock (1987) which found that the Union Paci®c±Missouri Paci®c (MP) merger resulted in little or no gain in market power since at least one other railroad served the same areas as UPMP. In contrast, an evaluation of the proposed SF±SP merger found large increases in ability to raise prices since no other railroad served southwest Kansas at that time. Thus this study and Lemke and Babcock (1987) conclude that mergers do not signi®cantly increase market power as long as some intrarail competition is maintained. However mergers that produce regional monopoly may result in substantial increases in market power.
This conclusion is supported by empirical evidence from other studies. For example Levin (1981) concluded that the intensity of interrailroad competition is a critical determinant of the level of railroad prices and pro®ts. He computed percent changes in the railroad prices for ®eld crops relative to the 1972 regulated base case. He estimated that with no interrailroad competition that these prices would increase by 40±80%. However he discovered that in the presence of a moderate degree of interrailroad competition that these prices would only increase by 5±6%. Thus as long as there is some interrailroad competition, increases in railroad post merger market power are likely to modest.
MacDonald (1989) estimated the eect of the Staggers Rail Act on railroad grain trans-portation. He found that prior to deregulation rail prices for Kansas wheat were relatively high due to lack of water carrier competition. However following deregulation he found that interrailroad competition produced 8±15% reductions in rail prices for Kansas wheat shipped to Texas Gulf of Mexico ports. The implication is that if interrailroad competition remains following railroad mergers, then increases in railroad market power are likely to be moderate.
7. Summary and conclusion
Table 8
Geographic price spreads prior to and after BN±SF and UP±SP mergers (cents per bushel)a
Elevator identi®er Pre-merger price spreadb Post-merger price spreadc Percent change between pre- and
has measured the maximum constraint on the ability of merged railroads to raise prices by as-suming that all rivals set prices at variable cost.
In subregion 1, the BN±SF merger produces an increase in net revenue and market power of the combined railroad and a reduction in net revenue and market power for both the UP and Kyle railroads. This result occurs because the cost of shipping wheat to Houston on the former BN is greatly reduced due to more direct routing. While BNSF net revenue and market power increase in subregion 1, the increases are not large because the BNSF cost advantage relative to the Kyle and the UP is relatively small.
In subregion 2 the BN±SF merger produces a redistribution of net revenue and market power from the Kyle to the BNSF. This is because the former BN is able to reduce its costs after the merger due to more direct routing. Also the former SF in subregion 2 assumes the lower costs of the BN after the merger. However the BNSF cost advantage relative to the other railroads serving subregion 2 is small and the combined railroad still faces a high degree of intramodal competition.
In subregion 3, the UP±SP merger produces an increase in net revenue and market power for the combined railroad. This can be attributed to two factors. First, the former SP routes to Houston are very circuitous so by merging with UP the former SP signi®cantly reduced its costs by using the more direct UP routes. Second, the assumption that the combined railroad will operate at the costs of the lower cost partner (UP) results in a decrease in former SP costs. The decline in costs of the combined railroad allows it to compete for more grain and enlarge its net revenue and market power. However, although UPSP gains an increase in net revenue and market power, the increases are not large since BNSF and CV also serve subregion 3, and the UPSP cost advantage relative to these railroads is not large.
In subregion 4 the BN±SF merger results in a modest net revenue increase for the combined railroad since the former SF is assumed to operate at the lower costs of the BN. Net revenue of the UPSP increases signi®cantly due to more direct routing of SP wheat shipments and the former SP is assumed to operate at the lower costs of the UP. The mergers produce virtually no change in market power for the railroads serving subregion 4. This is likely due to strong intermodal Table 8 (Continued)
Elevator identi®er Pre-merger price spreadb Post-merger price spreadc Percent change between pre- and
post-merger price spread
Sources: Data Transmission Network (DTN), Omaha, Nebraska provided wheat price data from the 52 Kansas el-evators. The Gulf of Mexico wheat price data is from U.S.D.A. Grain and Feed Weekly Summary and Statistics. bAverage price spread for the second quarter, 1994.
c
competition from commercial trucks which have a cost advantage for short hauls to terminal elevators at Wichita and Hutchinson, Kansas as well as Enid, Oklahoma.
Proponents of railroad mergers argue that mergers result in economies of scale and thus a reduction in transport cost per unit shipped. This study is unable to evaluate this argument since the ICC Railroad Costing Program is based on the assumption that railroads are a constant cost industry with constant and equal average and marginal cost. The paper documents signi®cant merger-related cost reductions but these reductions are attributable to factors such as more direct routing rather than economies of scale.
Finally, the study found that competitive rail prices will occur if shippers have access to at least two railroads. Data on actual rail prices are not available but can be estimated by comparing geographic price spreads for wheat between Kansas origins and Gulf destination. This is done for periods before and after the rail mergers. The data show that price spreads increased on only eight of 52 price spreads examined, (only two rose by more than 5%), and decreased on the remaining 44 price spreads. This shows that rail wheat transport prices from Kansas to the Gulf tended to decline after the rail mergers, consistent with the results of the network model.
References
Adam, B.D., Anderson, D.G., 1985. Implications of the Staggers Rail Act of 1980 for level and variability of Country Elevator price bids. Proceedings of the Transportation Research Forum 26, 357±363.
Babcock, M.W., 1981. Potential impact of rail deregulation in the Kansas wheat market. Journal of Economics 7, 93± 98.
Babcock, M.W., 1984. Eciency and Adjustment: The Impact of Railroad Deregulation. Cato Institute, Washington, DC.
Babcock, M.W., Sorenson, L.O., Chow, M.H., Klindworth, K.A., 1985. Impact of the Staggers Rail Act on agriculture: a Kansas case study. Proceedings of the Transportation Research Forum 26, 364±372.
Berwick, M., 1997. Truck Costs for Owner/Operators, MPC Report No. 97-81. Mountain-Plains Consortium, Fargo, North Dakota, September.
Casavant, K.L., 1985. An examination of the relationship between the Staggers Rail Act and agriculture in the United States. Proceedings of the Transportation Research Forum 26, 373±376.
Chow, M.H., 1984. Economic impacts of structural changes in the wheat logistics system: exporting winter wheat from twelve counties in Northeast Kansas. Ph.D. dissertation, Kansas State University, unpublished.
Dooley, F.J., 1991. Economies of size and density for short line railroads. MPC Report No 91-2, Upper Great Plains Transportation Institute, North Dakota State University, Fargo, ND.
Fuller, S.W., Makus, L., Taylor, M., 1983. Eect of railroad deregulation on export grain rates. North Central Journal of Agricultural Economics 5, 151±163.
Fuller, S.W., Bessler, D., MacDonald, J., Wolgenant, M., 1987. Eect of deregulation on export-grain rail rates in the plains and corn belt. Journal of the Transportation Research Forum 28, 160±167.
German, H.W., Babcock, M.W., 1982. The impact of rail deregulation on the movement of fresh fruits and vegetables. Logistics and Transportation Review 18, 373±384.
Hauser, R.J., 1986. Competitive forces in the US inland grain transportation industry: a regional perspective. The Logistics and Transportation Review 22, 158±173.
Interstate Commerce Commission, Bureau of Accounts. 1991 Uniform Railroad Costing System: Phase III Movement Costing Program Users Manuel, Washington, DC.
Kansas Department of Agriculture, 1993. Kansas Grain Marketing and Transportation: Data for 1992 Crop and Historical Data 1983±91, Topeka, Kansas.
Kansas Department of Agriculture, 1996. Farm Facts, Topeka, Kansas. Kansas Department of Agriculture, 1995. Farm Facts, Topeka, Kansas.
Kansas Department of Agriculture, 1994. Custom Wheat Cutter Rates, Topeka, Kansas, p. 9.
Kansas Grain and Feed Association, 1994. 1994 Kansas Ocial Guide, A Reference for Buyers and Sellers, Topeka, Kansas.
Klindworth, Keith, A., Sorenson, L.O., Babcock, M.W., Chow, M.H., 1985. Impacts of Rail Deregulation on Marketing of Kansas Wheat, US Department of Agriculture, Oce of Transportation, Washington, DC. Kwon, Y., Babcock, M.W, Sorenson, L.O., 1994. Railroad dierential pricing in unregulated transportation markets: a
Kansas case study. The Logistics and Transportation Review 30, 223±244.
Lemke, K., Babcock, M.W, 1987. Impact of rail mergers on Kansas export wheat rail rates and costs. Journal of the Transportation Research Forum 28, 151±159.
Levin, R.C., 1981. Railroad rates, pro®tability, and welfare under deregulation. Bell Journal of Economics 12, 1±20. MacDonald, J.M., 1987. Competition and rail rates for the shipment of corn, soybeans, and wheat. Rand Journal of
Economics 18, 151±163.
MacDonald, J.M., 1989. Railroad deregulation, innovation and competition: eects of the Staggers Act on grain transportation. Journal of Law and Economics 32, 63±95.
Park, J.J., 1998. The economic impact of rail mergers on Kansas export wheat rail rates. Ph.D. dissertation, Kansas State University, unpublished.
Sarwar, G., Anderson, D.G., 1986. Impact of the Staggers Act on variability and uncertainty of farm product prices. Proceedings of the Transportation Research Forum 27, 144±150.
Schmitz, J., Fuller, S.W., 1995. Eect of contract disclosure on railroad grain rates: an analysis of corn belt corridors. The Logistics and Transportation Review 3, 97±124.
Sorenson, L.O., 1984. Some impacts of rail regulatory changes on grain industries. American Journal of Agriculture Economics 66, 645±650.
Thompson, S.R., Hauser, R., Coughlin, B., 1990. The competitiveness of rail rates on export-bound grain. The Logistics and Transportation Review 26, 35±52.
US Department of Agriculture, 1982. Oce of Transportation. An Assessment of Impacts on Agriculture of the Staggers Rail Act and Motor Carrier Act of 1980, Washington, DC.
US Department of Agriculture, 1984. Eects of the Staggers Act on Grain Marketing, Washington, DC.
Werden, G.J., Froeb, L.M., 1994. The eects of mergers in dierentiated products industries: logit demand and merger policy. Journal of Law, Economics, and Organization 10 (2), 407±426.
Werden, G.J., Froeb, L.M., 1996. Simulation as an alternative to structural merger policy in dierentiated products industries. In: Coate, M.B., Kliet, A.N., Boston, M.A., (Eds.), The Economics of Antitrust. Kluwer Academic Publishers, Dordrecht, pp. 65±88.
Wilson, W.W., 1985. Railroad pricing and deregulation. Proceedings of the Transportation Research Forum 26, 105± 109.
