Andy Laurent and I share several common modeling interests; one of them is that we both are modeling harbor areas. Andy's new layout, set in 1952, features the Port of Green Bay, while mine, set in 1957, has the Port of Milwaukee as its setting. So when Andy asked whether there was a way to estimate the rail-delivered commodity flows of grain through an export elevator the question caught my attention.
Approach
It seemed reasonable to us to assume that the distribution of grains shipped by rail and terminating at an export elevator in Wisconsin could be modeled to first order by the distribution of grains shipped by rail that terminated in Wisconsin. So I proposed to look at the 1% Carload Waybill results for the appropriate grain commodity classes and find all the carloads in 1952 and 1957 that terminated in Wisconsin. After applying a few secondary adjustments (explained in detail below) this post will present the results of the analysis. In a follow up post, we'll compare our estimates to statistics kept by the Ports that date from the 1960s.
Why Bother?
First we want to address the question "Why bother?". If your primary interest is in delivering 40' box cars of "grain" to the elevator siding and flipping car cards, and if your operators "never read that part of the waybill", then this may be way too much detail for you. It is certainly not necessary to have this level of information to set up an operating session. But we would suggest that there are some considerations related to the types of grain shipped to elevators that might add operational interest:
![]()
![]()
Approach
It seemed reasonable to us to assume that the distribution of grains shipped by rail and terminating at an export elevator in Wisconsin could be modeled to first order by the distribution of grains shipped by rail that terminated in Wisconsin. So I proposed to look at the 1% Carload Waybill results for the appropriate grain commodity classes and find all the carloads in 1952 and 1957 that terminated in Wisconsin. After applying a few secondary adjustments (explained in detail below) this post will present the results of the analysis. In a follow up post, we'll compare our estimates to statistics kept by the Ports that date from the 1960s.
Why Bother?
First we want to address the question "Why bother?". If your primary interest is in delivering 40' box cars of "grain" to the elevator siding and flipping car cards, and if your operators "never read that part of the waybill", then this may be way too much detail for you. It is certainly not necessary to have this level of information to set up an operating session. But we would suggest that there are some considerations related to the types of grain shipped to elevators that might add operational interest:
- The different types of grain may be shipped in different types of freight cars. It turns out that this isn't the case for the 1950s (see detailed discussion below), but it is certainly possible that in later times as the use of covered hoppers increased, that the distribution of car types observed at an export elevator was a function of the types of grains shipped.
- The different types of grain may arrive at the export elevator at different times. The "grain rush" wasn't a single distinct period of time each and every year. If you want to use commodity flows as a time marker to add a distinct flavor to your operational setting, then it seems reasonable to consider the types of grain in a seasonal context.
- The different types of grain may arrive at the export elevator from different origins. For my setting, there are 9 different general routes by which freight cars arrive at the Port. Each of these routes has a different time of arrival and blocking scheme (order of the cars) as the cars appear on the layout from staging.
Besides all that, I think its just fun to see what sort of questions can be answered by the 1% Carload Waybill Sample, and I enjoy the process of reconstructing historical commodity flows.
The Commodity Classes
The commodity classes that might be associated with an export elevator are fairly self explanatory. The classes we investigated are:
- 001 - Wheat
- 003 - Corn
- 005 - Sorghum Grain
- 007 - Oats
- 009 - Barley and Rye
- 011 - Rice
- 013 - Grain N.O.S.
- 043 - Soybeans
Sorghum grain was used in the 1950s as a ingredient in animal feed and to make molasses. It was primarily grown in the southeastern U. S. The commodity class Grain N.O.S. is a catch all class (N.O.S. means Not Otherwise Specified). The commodities listed by the ICC in this class are buckwheat, popcorn, spelt, and grain noibn (noibn means not otherwise indexed by name).
National Statistics
First we constructed a table from the 1% Carload Samples from 1947 through 1960 to identify which of these commodity classes were the important actors and whether there were any gross temporal trends. This chart shows the data (click to enlarge or download):
Clearly on a national basis the commodity flows of grains were dominated by wheat and corn. An analysis of the data shows considerable variation about the average of the 14-year period. In all cases, the yearly variation (the row labeled Std Dev is an estimate of the variation from year to year) is much greater than the uncertainty in the sample (the row labeled Uncert is an estimate of the sampling error). So these temporal changes are real, not artifacts of the survey or sampling methodology, and can be used as time markers. They reflect good/bad harvests, changes in rail versus freight traffic, differences in export networks and production, etc.
| Year | Wheat | Corn | Sorghum | Oats | Barley | Rice | Grain NOS | Soybeans |
| 1947 | 8,559 | 4,679 | 516 | 1,500 | 1,455 | 297 | 72 | 906 |
| 1948 | 7,971 | 3,096 | 419 | 1,153 | 1,170 | 288 | 50 | 983 |
| 1949 | 7,077 | 3,853 | 532 | 1,103 | 1,321 | 335 | 43 | 1,249 |
| 1950 | 5,437 | 3,262 | 814 | 878 | 1,017 | 340 | 46 | 989 |
| 1951 | 6,799 | 3,660 | 1,131 | 865 | 1,073 | 374 | 44 | 1,219 |
| 1952 | 6,635 | 3,604 | 524 | 861 | 971 | 459 | 34 | 1,252 |
| 1953 | 5,965 | 3,626 | 197 | 898 | 833 | 435 | 42 | 1,227 |
| 1954 | 5,914 | 3,349 | 585 | 885 | 1,012 | 400 | 34 | 1,143 |
| 1955 | 5,656 | 3,407 | 661 | 924 | 1,257 | 536 | 37 | 1,425 |
| 1956 | 6,609 | 3,491 | 595 | 866 | 1,514 | 589 | 36 | 1,307 |
| 1957 | 5,790 | 4,168 | 906 | 714 | 1,292 | 405 | 24 | 1,387 |
| 1958 | 6,434 | 3,962 | 1,931 | 786 | 1,586 | 324 | 28 | 1,472 |
| 1959 | 5,922 | 3,772 | 1,418 | 738 | 1,241 | 340 | 13 | 1,600 |
| 1960 | 6,646 | 3,777 | 1,639 | 657 | 1,048 | 372 | 18 | 1,628 |
| Avg | 6,530 | 3,693 | 848 | 916 | 1,199 | 392 | 37 | 1,271 |
| Std Dev | 885 | 402 | 504 | 215 | 221 | 88 | 15 | 221 |
| Uncert | 81 | 61 | 29 | 30 | 35 | 20 | 6 | 36 |
Freight Car Types
The next two tables show the distribution of freight car types for the different commodity classes. This table is for 1952:
| Car | Wheat | Corn | Sorghum | Oats | Barley | Rice | Grain NOS | Soybeans |
| Box | 6,621 | 3,634 | 522 | 870 | 980 | 457 | 34 | 1,258 |
| Refrig | 1 | 0 | 0 | 0 | 0 | 1 | 0 | 0 |
| Stock | 11 | 4 | 5 | 0 | 0 | 0 | 0 | 0 |
| Gondola | 7 | 2 | 0 | 0 | 0 | 0 | 0 | 0 |
| Hopper | 23 | 9 | 0 | 0 | 0 | 1 | 0 | 5 |
| Special | 28 | 3 | 0 | 0 | 0 | 3 | 0 | 1 |
| %Box | 98.95% | 99.51% | 99.05% | 100% | 100% | 98.92% | 100% | 99.53% |
And this table is for 1957:
| Car | Wheat | Corn | Sorghum | Oats | Barley | Rice | Grain NOS | Soybeans |
| Box | 5,767 | 4,158 | 886 | 712 | 1,282 | 377 | 23 | 1,386 |
| Refrig | 0 | 0 | 0 | 0 | 0 | 1 | 1 | 0 |
| Stock | 1 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| Gondola | 0 | 1 | 0 | 0 | 0 | 0 | 0 | 0 |
| Hopper | 6 | 3 | 3 | 0 | 6 | 9 | 0 | 1 |
| Flat | 0 | 0 | 0 | 0 | 0 | 1 | 0 | 0 |
| Special | 16 | 6 | 17 | 2 | 4 | 17 | 0 | 0 |
| %Box | 99.60% | 99.76% | 97.79% | 100% | 99% | 93.09% | 96% | 99.93% |
For all of these commodity classes, at least 99 out of every 100 shipments were made in box cars. This result is not time dependent. I looked at the data from 1960 as well and the results were the same. The transition from shipping grain in box cars to shipping in covered hoppers (which would show up as Special in the tables above) post-dated 1960.
Seasonality
The quarterly summaries that were prepared by the ICC cover the period 1947 through 1952. I found the results interesting as there are several different types of behavior as can be seen from this graph:
Wheat shipments show a strong seasonal peak that is fairly regular. Corn shipments show a much more irregular pattern, the peak widths are broader, and the peak heights are not nearly as pronounced. Oat shipments hardly have a seasonal variation at all. These results suggest that in the late 1940s and early 1950s the grain rush was not a short distinct event that occurred with regularity each year. There was a more complex pattern to the commodity flows as the out-of-phase waves constructively and destructively interfered.
The State to State Flows - First Order Solution
We next looked at the state to state flows, selecting the commodity flows from any state that terminated in Wisconsin. This table shows the raw data for 1952:
| To Wisconsin | |||||||
| From: | Wheat | Corn | Oats | Barley | Rice | Grain NOS | Soybeans |
| California | 0 | 0 | 0 | 1 | 0 | 0 | 0 |
| Colorado | 1 | 0 | 0 | 0 | 0 | 0 | 0 |
| Idaho | 0 | 1 | 0 | 6 | 0 | 0 | 0 |
| Illinois | 0 | 18 | 1 | 0 | 1 | 0 | 0 |
| Indiana | 4 | 0 | 0 | 0 | 0 | 0 | 0 |
| Iowa | 1 | 17 | 1 | 1 | 0 | 0 | 0 |
| Kansas | 5 | 0 | 0 | 0 | 0 | 0 | 0 |
| Louisiana | 0 | 0 | 0 | 0 | 5 | 0 | 0 |
| Michigan | 0 | 0 | 0 | 1 | 0 | 0 | 0 |
| Minnesota | 25 | 40 | 22 | 162 | 0 | 2 | 4 |
| Montana | 63 | 0 | 0 | 0 | 0 | 0 | 0 |
| Nebraska | 6 | 0 | 0 | 1 | 0 | 0 | 0 |
| North Dakota | 188 | 0 | 1 | 9 | 0 | 0 | 0 |
| Oregon | 0 | 0 | 0 | 9 | 0 | 0 | 0 |
| South Dakota | 14 | 2 | 0 | 2 | 0 | 0 | 0 |
| Texas | 0 | 0 | 0 | 0 | 2 | 0 | 0 |
| Washington | 0 | 0 | 0 | 4 | 0 | 0 | 0 |
| Wisconsin | 1 | 5 | 4 | 19 | 0 | 1 | 0 |
| Total | 308 | 83 | 29 | 214 | 8 | 3 | 4 |
| 47.5% | 12.8% | 4.5% | 33.0% | 1.2% | 0.5% | 0.6% |
And this is the corresponding table for 1957:
| To Wisconsin | |||||||
| From: | Wheat | Corn | Oats | Barley | Rice | Grain NOS | Soybeans |
| Arkansas | 0 | 0 | 0 | 0 | 3 | 0 | 0 |
| California | 0 | 0 | 0 | 0 | 0 | 1 | 0 |
| Colorado | 0 | 0 | 0 | 0 | 0 | 1 | 0 |
| Idaho | 0 | 0 | 0 | 5 | 0 | 0 | 0 |
| Illinois | 0 | 20 | 0 | 0 | 0 | 9 | 0 |
| Indiana | 0 | 2 | 0 | 0 | 0 | 0 | 0 |
| Iowa | 0 | 28 | 39 | 0 | 0 | 0 | 1 |
| Kansas | 0 | 0 | 0 | 0 | 0 | 3 | 0 |
| Louisiana | 0 | 0 | 0 | 0 | 5 | 0 | 0 |
| Michigan | 0 | 0 | 0 | 0 | 0 | 4 | 0 |
| Minnesota | 27 | 4 | 15 | 136 | 0 | 50 | 0 |
| Missouri | 0 | 0 | 0 | 0 | 0 | 3 | 0 |
| Montana | 25 | 0 | 0 | 1 | 0 | 0 | 0 |
| Nebraska | 1 | 0 | 0 | 0 | 0 | 1 | 0 |
| North Dakota | 132 | 1 | 8 | 29 | 0 | 3 | 0 |
| Oregon | 0 | 0 | 0 | 7 | 0 | 0 | 0 |
| South Dakota | 5 | 0 | 0 | 0 | 0 | 0 | 0 |
| Texas | 0 | 0 | 0 | 0 | 3 | 0 | 0 |
| Wisconsin | 1 | 14 | 6 | 31 | 0 | 0 | 0 |
| Total | 191 | 69 | 68 | 209 | 8 | 74 | 1 |
| 30.8% | 11.1% | 11.0% | 33.7% | 1.3% | 11.9% | 0.2% |
These results show that shipments of wheat made up between a third and one half of the grain commodity flows to Wisconsin, that barley made up another third, and that the remainder was scattered between corn, oats, and other grains (Grain N.O.S.). They also shows that North Dakota and Minnesota were important shipping states. Note the rise in Grain N.O.S. shipments (primarily from Minnesota) in 1957 relative to 1952 - this rise is greater than the sampling error and seems to represent a one-year event rather than a trend.
Second Order Corrections
Note that in both of these tables the rows for Illinois, Indiana, and Iowa are highlighted yellow. Each of these states had a well developed export network and it seems unlikely that a lot of this rail traffic was destined for export elevators in Wisconsin. Similarly, the columns for barley and rice are highlighted in yellow. Wisconsin had a high demand for barley (to make malt) and rice to support its brewing industry. It also seems likely that a lot of these two commodity flows were to support in-state consumptive uses rather than for export. So we reduced the commodity flows from those three states and from those two commodity classes by 90% (a somewhat arbitrary number) to come up with this estimate for 1952:
| To Wisconsin | |||||||
| From: | Wheat | Corn | Oats | Barley | Rice | Grain NOS | Soybeans |
| California | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| Colorado | 1 | 0 | 0 | 0 | 0 | 0 | 0 |
| Idaho | 0 | 1 | 0 | 1 | 0 | 0 | 0 |
| Illinois | 0 | 2 | 0 | 0 | 0 | 0 | 0 |
| Indiana | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| Iowa | 0 | 2 | 0 | 0 | 0 | 0 | 0 |
| Kansas | 5 | 0 | 0 | 0 | 0 | 0 | 0 |
| Louisiana | 0 | 0 | 0 | 0 | 1 | 0 | 0 |
| Michigan | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| Minnesota | 25 | 40 | 22 | 16 | 0 | 2 | 4 |
| Montana | 63 | 0 | 0 | 0 | 0 | 0 | 0 |
| Nebraska | 6 | 0 | 0 | 0 | 0 | 0 | 0 |
| North Dakota | 188 | 0 | 1 | 1 | 0 | 0 | 0 |
| Oregon | 0 | 0 | 0 | 1 | 0 | 0 | 0 |
| South Dakota | 14 | 2 | 0 | 0 | 0 | 0 | 0 |
| Texas | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| Washington | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| Wisconsin | 1 | 5 | 4 | 2 | 0 | 1 | 0 |
| Total | 304 | 52 | 27 | 21 | 1 | 3 | 4 |
| 73.8% | 12.5% | 6.6% | 5.2% | 0.2% | 0.7% | 1.0% |
and this estimate for 1957:
| To Wisconsin | |||||||
| From: | Wheat | Corn | Oats | Barley | Rice | Grain NOS | Soybeans |
| Arkansas | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| California | 0 | 0 | 0 | 0 | 0 | 1 | 0 |
| Colorado | 0 | 0 | 0 | 0 | 0 | 1 | 0 |
| Idaho | 0 | 0 | 0 | 1 | 0 | 0 | 0 |
| Illinois | 0 | 2 | 0 | 0 | 0 | 1 | 0 |
| Indiana | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| Iowa | 0 | 3 | 4 | 0 | 0 | 0 | 0 |
| Kansas | 0 | 0 | 0 | 0 | 0 | 3 | 0 |
| Louisiana | 0 | 0 | 0 | 0 | 1 | 0 | 0 |
| Michigan | 0 | 0 | 0 | 0 | 0 | 4 | 0 |
| Minnesota | 27 | 4 | 15 | 14 | 0 | 50 | 0 |
| Missouri | 0 | 0 | 0 | 0 | 0 | 3 | 0 |
| Montana | 25 | 0 | 0 | 0 | 0 | 0 | 0 |
| Nebraska | 1 | 0 | 0 | 0 | 0 | 1 | 0 |
| North Dakota | 132 | 1 | 8 | 3 | 0 | 3 | 0 |
| Oregon | 0 | 0 | 0 | 1 | 0 | 0 | 0 |
| South Dakota | 5 | 0 | 0 | 0 | 0 | 0 | 0 |
| Texas | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| Wisconsin | 1 | 14 | 6 | 3 | 0 | 0 | 0 |
| Total | 191 | 24 | 33 | 21 | 1 | 66 | 0 |
| 56.9% | 7.2% | 9.8% | 6.2% | 0.2% | 19.6% | 0.0% |
Summary
- On a national basis during the late 1940s and 1950s, the predominant commodity flows in grain were comprised of wheat and corn. Soybean shipments showed a secular upward trend during this time. Yearly commodity flows for all grains were moderately variable about the mean.
- More than 99 grain shipments out of 100 were in box cars.
- There was a seasonality to grain shipments, but the seasonality wasn't characterized by a simple regular peak that occurred at the same time each year.
- For export elevators in Wisconsin in 1952, our model suggests that at least 75% of the shipments of grain were wheat, about 12% were corn, and the rest scattered among several commodity classes. A lot of the traffic came from North Dakota, Montana, and Minnesota.
- For export elevators in Wisconsin in 1957, our model suggests that about 60% of the shipments of grain were wheat, about 20% Grain N.O.S., and the remainder scatted among several commodity classes. A lot of the traffic came from North Dakota, Minnesota, and Montana.
Charles Hostetler and Andy Laurent