5.2 COMMODITY FLOW AND ACCESS TO MARKETS

[Table of Contents] [Previous Chapter] [Next Chapter] [Main Page]

[Next Section] [Previous Section] [Chapter Head]

Adams, William H., Peter M. Bowers, and Catherine M. Williams. (1998) 5.2 Commodity Flow and Access to Markets. In Historical Development of the Chena River Waterfront, Fairbanks, Alaska: An Archaeological Perspective, edited and compiled by Peter M. Bowers and Brian L. Gannon, CD-ROM. Alaska Department of Transportation and Public Facilities, Fairbanks.

Tying artifacts to their sources in the national or international marketplace can provide useful information about how those artifacts got to the local site, and how they fit into the system of production and distribution. The gold rush era found men and women in what was popularly considered the "ends of the Earth," struggling to overcome hardships and achieve success. While distances to the Continental United States were great and physical obstacles such as mountains and rivers were substantial, analysis of the Barnette archaeological collection indicates that people living in Fairbanks experienced only a few short years when they did not have access to virtually anything someone in Seattle could buy. A commodity took a few more weeks or months to arrive in Fairbanks and cost a great deal more, but if it could be found on the Pacific West Coast, it was essentially available in Fairbanks.

The Geographer’s Commodity Flow Model

Geographers have focused on the way the national economy produces and distributes manufactured goods. They have viewed how these goods travel from the manufacturer to the consumer, a process called commodity flow. In 1964, Allan Pred devised a typology for commodity flows using industry type and market accessibility.¹²

Industry type was divided by Pred into three groups. The Raw Material and Power Oriented Industries focus on extracting raw materials and refining them for transport elsewhere to be made into finished products; hence, these tend to be located away from population areas. The second group are Market Oriented Industries; these companies serve the regional and national markets, and hence need to be located in areas with good market accessibility. The third industry, Labor Related or Agglomeration Economies, involve products with a high enough demand, or low enough production cost, that transportation costs are not a factor in the product’s sales.

Pred defined accessibility using New York City as the baseline, and then ranking different areas of the country in relation to New York City in terms of access through a combined land and sea transportation network. He arbitrarily defined these areas as being high access (0-25% less than New York City), intermediate (25-40% less), and low (more than 40%; Figure 5.3). Accessibility is largely a function of the development of the transportation web and the population being served by that network. Clearly, the larger the population, the greater the need for transporting goods to those people, and hence the greater access they have to manufactured goods. Put another way, people in large cities have greater access to material goods than people in rural areas, and people in areas of the country with lots of large cities have even more access. This means they can get a greater variety of material goods at cheaper prices. Small town stores cannot afford to bring in a huge variety of merchandise.

Figure 5.3. Market Access Areas (redrawn from Pred 1970).

Pred found that Market Oriented manufacturers located in high market access areas distributed their goods the furthest, with the high access area getting the largest percentage, but the intermediate and low access areas also receiving a share. The same kind of manufacturers located in intermediate access areas had a smaller distribution network, mostly sent short distances and within the intermediate access areas, for they could not compete in the high access or low access areas.¹³

The Archaeologist’s Commodity Flow Model

The geographer’s model for commodity flow traced the manufactured goods to the consumer and provided a view of the consumer’s accessibility to manufactured goods. The archaeological data look at this backwards. We see the end result of consumerism--disposal--and interpret from these discarded objects what their relationship was in terms of manufacture and distribution. Because these are mirrored images in a sense, they should reveal the same information. The archaeological data are seen through the filter of time and disposal practices, but the broad patterns of commodity flows and consumer choices should still be apparent.

A few historical archaeologists began the quest for economic models of market access in the mid-1970s. Joel Klein addressed the topic of changing market economies theoretically, but did not present any hypotheses or data.¹⁴ Robert Schuyler also examined the issue of trading networks in his study of Sandy Ground, New York.¹⁵ However, the first extensive treatment was William H. Adams’ study of Silcott, Washington.¹⁶ This study of seven sites dating from the 1900-30 period laid the groundwork for future studies like the Barnette Project by incorporating not only archival data, but also oral history in an integrative way. In the Silcott study, Adams identified the manufacturing locations for 1,043 artifacts recovered from the seven sites and found that 87.8% of the identified products came from more than 1,000 miles away. They were made in the American Manufacturing Belt, the northeastern quarter of the United States, where 65% of the US manufacturing capacity was located at the turn of the century.¹⁷

Part of the explanation for this is obvious; since most things were made in that Manufacturing Belt, then most objects found in an American site will come from that region. The commodity flow thesis also posits, however, that West Coast manufacturers should have made their presence felt more clearly in the Silcott archaeological assemblages, whereas those western-made artifacts were relatively uncommon. Timothy B. Riordan and Adams used data from two communities in Mississippi, Waverly Plantation and Bay Springs, to further test of the model (see summary below).¹⁸ The thesis generated is that "when located in different geographic regions, sites having the same access to the national market will show greater similarity to each other than to sites having different access, even when located in the same region."¹⁹

Study Summaries. This section provides summaries of four different commodity flow studies that have been carried out. It illustrates the utility of the model and provides the background for comparison with the Barnette data analysis.

Waverly Plantation, Bay Springs, Silcott, and Sandy Ground. In 1979, Adams and Riordan investigated Waverly Plantation and Bay Springs, Mississippi. The Waverly Plantation sites were four tenant farmer houses occupied by African-Americans in the 1880-1930 period; one site was also used by white sharecroppers from 1930-50.²⁰ The Bay Springs sites date generally from 1840-90 and were associated with white mill workers.²¹ These two communities were located in the intermediate access area defined by Allan Pred. For comparison sake, two other communities were selected: Sandy Ground and Silcott. Silcott was a white farming community in the low access area of southeastern Washington with sites dated primarily from 1900-30. Sandy Ground is near New York City, and so lies in the heart of the high access area. The archaeological assemblage there dated 1890-1930.²² To compare these assemblages, artifact percentages were used in the analysis.

Analysis of the assemblages from these sites supported Pred’s thesis. The site found in the high access area, Sandy Ground, derived 99.4% of its material from manufacturers also located in the high access area while only 0.6% came from low access area manufacturers (Table 5.4). Silcott, located in a low access area, obtained 56.5% of its manufactured goods from the high access area manufacturers, while manufacturers located in low and intermediate access areas provided Silcott consumers with 23.0% and 20.5%, respectively, of their materials. Waverly and Bay Springs, located in intermediate access area, received 32.5% and 28.6% of their goods from manufacturers in the intermediate access area; and negligible quantities from low access manufacturers. What these data suggest is that the majority of manufactured goods come from manufacturers located in the high access area, which is also the area that produces 65% of the manufactured goods for the United States. The results range from 56.5% to 71.4% for Silcott, Waverly, and Bay Springs and support Pred’s model. Pred also postulated that companies located in intermediate areas tended to distribute and be competitive only in those areas. This is supported by the data from Waverly and Bay Springs, which had the highest percentages for manufacturers located in the intermediate area. The Silcott data indicate that manufacturers located in low access areas also compete best in those areas. What all these data show is that each community purchased goods made close to home when these were available.

Table 5.4. Market Oriented Artifacts from the Original Model.

Do the same observations hold true for Labor Related artifacts? These kinds of goods are ones either produced inexpensively or had so much value added that transportation was not a factor. Hence, we should expect that goods manufactured in the high access area should be able to out compete goods made anywhere else. Based on this relatively small sample, the expectation seems to be met (Table 5.5). All these kinds of goods found at Silcott and Bay Springs were made in the high access area, while only at Waverly were goods manufactured in low and intermediate areas found. The ten artifacts from Waverly made in a low access area came from a single manufacturer of tax tokens in Denver.

Table 5.5. Labor Related Artifacts from the Original Model.

Ashley Plantation. Richard D. Brooks used the commodity flow model at Ashley Plantation, located on the Savannah River in South Carolina. The sites dated from 1876-1950, and consisted of three tenant farmer houses, two dwellings, and a mill complex.²³ This plantation is located virtually the same distance as Waverly Plantation was from New York City in terms of the commodity flow model, roughly at the 37% cline of market access. Hence, one would expect that the figures should be comparable, and they are. Market Oriented goods from the high access area at Ashley Plantation represented 73.8% (compared to 66.0% at Waverly), while 23.2% and 3.0% came from intermediate and low access area manufacturers respectively, compared to 32.5% and 1.5% for Waverly (Table 5.6).

Table 5.6. Market Oriented vs. Labor Related Artifacts from Ashley Plantation, South Carolina.

Oregon. Lou Ann Speulda and Gary C. Bowyer²⁴ have tested the model on five sites in Oregon: Homestead site JE-3 dated circa 1906-13, Homestead site JE-2 dated circa 1910-35, Homestead site JE-1 dated circa 1923-38, the logging camp circa 1922-31, and the railroad construction camp circa 1923-26 (Table 5.7). These sites are located in the 50-55% cline within the low access area. Based upon the results from Silcott, which is located in the same region, we would expect that the high access area figures to be about 57%. The homesteads have very similar frequencies to Silcott for high access artifacts, while the range for intermediate access area manufacturers is rather low. These results support the contention that sites located in low access areas will have higher representation of low access manufacturers than intermediate access ones. The logging camp and the railroad construction camp data clearly show that we must be careful to identify the kind of sites being analyzed, and that we need to control for biases introduced by the kind of population occupying these sites. The railroad camp was male dominated while the logging camp had families. This logging camp was moved en masse every few years. Its occupants were not limited to the company store for purchases. Three root cellars, some disposal pits, and habitation areas were examined there. The railroad camp was a very short term occupation; trash pits were excavated there. What this study shows is that site type can also affect market access. Specialized or industrial sites should not be compared to residential sites, but each type must be compared with each other.

Table 5.7. Market Oriented Artifacts from Oregon Sites.

* 103 of these were tobacco tins. Eliminating these, the percentages become: low 45.2%, intermediate 19.3%, and high 35.5%.

Peoria, Illinois. In each of the studies presented so far, the data have been lumped for all the sites in a community and without consideration of chronology. When the period represented by an artifact collection cross-cuts major changes in the transportation network, a more accurate picture would be obtained by separating the assemblage into chronological periods. Certainly the access to the national, regional, and even local market will differ as transportation improves, so dividing it into periods based upon transportation history makes sense.

Recognizing the value of using chronological periods, Melanie A. Cabak and Mark D. Groover conducted an analysis of an artifact assemblage from Peoria, Illinois in 1993.²⁵ Peoria is located on the fringe of the high access area in the 24% cline (Table 5.8). They did not distinguish their artifacts on the basis of industry as had been done by earlier studies. What they were able to suggest, albeit with a relatively small sample size, was that the high access area manufacturers dominated early on, but through time the manufacturers in the intermediate area steadily increased their market share. Clearly, these data suggest that we should control closely for time and that the closer we get to the present, the more the intermediate and low access manufacturers will have had a chance to develop their industries. Another factor which may need to be considered is the regional history of industrial development. Different areas of the country have developed at different rates, as has the national transportation network that links them together.

Table 5.8. Market Access Over Time: an Example from Peoria, Illinois.

The Barnette Analysis

Alaska was not included in Pred’s original access area model, so we are proposing here that Alaska was comparable to low access areas in the Continental United States (Figure 5.3).²⁶ While Alaskan sites were in the low access area and would be expected to be most similar to each other, we would also expect to see quantifiable similarities with collections from non-Alaskan sites in low market access areas.

What is the situation for the sites in Fairbanks? Did they draw, like most low-access area sites, upon the main industrial region of the United States for their supplies, or did West Coast manufacturers provide more of the goods? Did Canadian manufacturers provide a larger amount of the supplies due to their proximity and to the extreme demand the influx of thousands of Stampeders had created? Did certain kinds of goods travel further than other kinds? Was there a difference in the supply of goods in Fairbanks between the "Steamboat era" (1901-22) and the "Railroad era" (1923-41)?²⁷ Is that difference represented in the archaeological assemblages?

Weighted Analysis. In our analysis of the Fairbanks assemblage from the Barnette Project, we have divided the collection into three time periods: Steamboat era, Railroad era, and Other (Tables 5.9 and 5.10; see also Section 2.2). These subassemblages combine the materials from different areas of the site based upon their association with deposits dating from the Steamboat era (1901-22) or the Railroad era (1923-41). Those artifacts which could not be definitely assigned to one or the other were placed in the Other category. Table 5.9 provides a summary of all the artifacts from the Barnette Project that contain trade- or manufacturer marks, are from a dateable stratigraphic context, and/or are assignable to a manufacture date or date range. In the case of artifacts that served as containers (e.g., bottles), we attempted to distinguish between where a container was manufactured as opposed to where it was filled. In this analysis, we have attempted to use the product manufacturer when possible.

View Table 5.9

The maps in Figures 5.4 and 5.5 are visual summaries of company locations for Steamboat era and Railroad era artifacts. The numbers in each state represent the number of identifiable places of manufacture of a Barnette collection artifact type which contains identifiable manufacturer or trademark information and is from a dated geologic context. They do not reflect quantities of artifacts recovered, but a simple count of different manufacture sources. These maps show the shift in manufacturer locations that took place between the two time periods. Although numbers of high access area commodities remained about the same, the numbers of West Coast (low access area) manufacturers increased in the later period. Surprisingly, the number of Alaskan manufacturers declined during the Railroad era.

Figure 5.4. Company locations for Barnette artifacts from the Steamboat era (1901-22).

Figure 5.5. Company locations for Barnette artifacts from the Railroad era (1923-41).

Using a comparison of artifact frequency by access area and time period, percentages of artifacts coming from the west coast manufacturers increased from 10.7% in the Steamboat era to 27.9% in the Railroad era. Intermediate access suppliers dropped from 42.8% to 14.2% between the early and late time period, and high access manufacturers increased from 46.5% to 57.9% over time (Table 5.10).

Table 5.10. Comparisons Using Frequency of Artifacts by Access Area for Different Time Periods.

Because of the very large number of beer bottles coming from St. Louis, Missouri breweries and because these bottles may have been refilled locally at the Barthel Brewery, we decided to also look at the assemblage without those two artifact types (Marks 339 and 340). We have adjusted the figures to remove these bottles from the assemblage (Table 5.11). By doing this, the contribution of the intermediate access area drops considerably to 16.4%, while the low access area is 23.4% and the high access share is 60.2%.

Table 5.11. Comparisons Using Frequency of Artifacts by Access Area, without Marks 339 and 340.

Eliminating the "Other" era from Table 5.10 provides a useful check on the method. Table 5.12 includes summaries and comparisons of all these "key" Barnette artifacts (N=603): those from Table 5.9 that had a dateable stratigraphic context. This table also illustrates that the high access area was the source for most goods during both periods, and its share increased between the early period and the later period. The representation of the intermediate and low access areas also changed from the Steamboat to the Railroad eras.

Table 5.12. Comparisons Using Frequency of Artifacts by Access Area for Steamboat and Railroad Time Periods.

There are several possible explanations for these changes. During the early "Steamboat era," water transportation was generally less expensive than rail transportation across the Continental United States.²⁸ The problem in Alaska is that for goods to come into the interior by water, they must travel a considerable distance along the Pacific coast, into the Bering Sea, and then all the way up the Yukon River from St. Michael to Fairbanks. When the railroad was completed in 1923 this distance was shortened greatly. As a result, rail transport could not only compete in price, but also offer shorter delivery time. The railroad was also government subsidized, which kept costs artificially low.

The second consideration is that by the mid 1920s, the West Coast manufacturers had an additional two decades of growth, in part fueled by the northern trade. Over time the West Coast manufacturers were able to capture a larger share of the Fairbanks market, due both to internal growth and improved transportation systems. West Coast manufacturers also geared their advertising and distribution networks very heavily toward the northern gold rush market. They even advertised and "boomed" the strikes to encourage trade. The study done of Peoria, Illinois (Table 5.8) showed a similar trend towards development of the "local" access area. In part, what both the Peoria and the Fairbanks assemblages are showing is that through time American industry developed further west to serve the local and regional markets.

Another useful comparison in the Barnette data is the percentage of foreign-manufactured goods which occurred between the Steamboat and Railroad eras. This is summarized in Table 5.13. Products from foreign sources stay relatively consistent in percentage as the railroad replaced the steamboats. Why this relationship stays the same is unclear, although it appears that domestic transportation networks had little effect on use of imported goods.

Table 5.13. Comparison of frequency of Foreign-made goods by time periods.

Unweighted Analysis. In each of the studies above, the samples are weighted by the numbers of artifacts each company produced which were found at the archaeological sites. We have also tried a different approach here, namely an unweighted sample in which we have counted each company represented in the archaeological assemblage, but have ignored the artifact totals (Table 5.14). Does this method yield acceptable results? The data suggest that this approach yields results comparable to the weighted analysis, and with much simpler calculations. Fairbanks relied upon companies in the low access area (West Coast primarily) much more than did the residents of Waverly Plantation, Bay Springs, or Ashley Plantation.

Of the 202 different identifiable companies represented in the Fairbanks artifact assemblage (Table 5.9), 61 or 30% came from the low access area, compared to between 0-4% for the sites located in the intermediate access area. In other words, if the approach is valid, companies located in low access areas compete best in the low access areas. Companies located in the intermediate access area represented 14% of the Fairbanks assemblage, and between 18% and 27% of the assemblages for intermediate access area sites (Waverly, Bay Springs, and Ashley Plantation). Companies from the high access area represented 56% of the Fairbanks assemblage companies, compared to a range of 70-79% for the three other assemblages in the intermediate access area.

Clearly the Fairbanks artifact assemblage is distinctly different from the ones located in the intermediate access area. This may be due in part to a bias in the archaeological record attributable to functional variation and artifact preservation at each of the saloon sites. However, the record indicates that merchants in Fairbanks were able to obtain more of their merchandise from low- and intermediate- area manufacturers than we might have expected. All other things being equal, this suggests that the shipping costs of getting goods from an East Coast manufacturer to Fairbanks were so significant that manufacturers from the high access areas could not always compete head-to-head with manufacturers located on the West Coast making the same products. A more detailed historical analysis of product histories would probably reveal that many products from high access area manufacturers were not made by any West Coast companies, which is why high access goods appeared in the Fairbanks assemblage at all.

Table 5.14. Unweighted Comparisons Using Frequency of Companies by Access Area.

The next step in our analysis was to evaluate the two methods. The original model proposed by Riordan and Adams requires that one establish the object count for every artifact. The other method illustrated with Table 5.14 is much simpler, for it uses an unweighted count of the number of companies. Are the results similar? We believe they are close enough that in collections with a large sample size like that from the Barnette Project, it may be possible to use this alternative method. The problem is, of course, that we lack comparable data from other sites to test its validity and refine its usefulness. However the overall story each method tells us is the same. With time, the companies and their products coming from low access areas were more important (in terms of relative numbers) than those coming from the intermediate access areas (Table 5.15). We believe this alternative method bears further investigation.

Table 5.15. Comparison of the entire Fairbanks Assemblage Using the Unweighted Total for Companies and the Weighted Total for Artifacts.

Let us now examine how the Fairbanks assemblage compares with those sites discussed earlier (Table 5.16 and 5.17). We have arranged these tables by project with those furthest from New York City on the left. We have combined numbers for market oriented and labor related goods as necessary. We have not considered the industrial sites from Oregon, and have assumed that the Barnette collection, although partially from commercial contexts, is still comparable to the residential sites used. The three groupings by access areas are strongly supportive of the original model proposed by Riordan and Adams. The sites found in Alaska, Washington, and Oregon, in the low access area, have remarkably similar profiles for their access to goods, with a range of 19.7% to 21.2% for low access area manufacturers, a range of 14.5% to 29.8% for intermediate access area manufactures, and a range of 50.5% to 64.4% for high access area manufacturers. The sites located in the intermediate access area in Mississippi and South Carolina range from 0% to 4.4% for low, 18.1% to 22.8% for intermediate, and 72.8% to 81.9% for high access area manufacturers. The data from Peoria, located on the very fringe of the high access area, indicated it is most comparable to the sites located in the intermediate access area. The one site purely in the high access area, Sandy Ground, shows a drastically different profile than the other site assemblages, with nearly all goods coming from that area.

Thus, the Pred²⁹ model appears to be a valid one. The economic geography model has been projected into the archaeological record and similar results have been derived. These data and observations reveal that consumers in the low access area used manufactured goods made in that same area much more commonly than those made in the intermediate access area. People living in the intermediate access area chose products made in that area, more than those made in the low access area. In both cases the respective regional manufacturers accounted for about 20% of the overall assemblages. In addition, the further the site is from New York City, the less common are goods made in the high access area.

Table 5.16. Comparison of Different Assemblages by Access Area (Artifact Totals).

Table 5.17. Comparison of Different Assemblages by Access Area (Percentages).

Conclusions

In this section we have compared the data from several projects with the Barnette Project assemblage in terms of the commodity flow model. We have found that the original model is a valid approach to understanding the relationship between consumer and manufacturer as revealed in the archaeological data. Archaeological assemblages in different areas of the country were found to have predictable relationships. More studies of this type will help clarify biases introduced by time, site function, and artifact preservation, and allow refinement of the clinal access areas (Figure 5.3) to further increase the utility of this model.

We deviated from the original model by lumping market oriented and labor related artifacts. The reason for this is that much published archaeological data cannot be split this way. We also tried a new approach, namely using an unweighted sample by simply totaling the numbers of companies from each area instead of totaling the numbers of artifacts by company. Because the results were comparable using both approaches, we recommend future studies use both approaches to see if the simpler method is valid.

[Next Section] [Previous Section] [Chapter Head]

[Table of Contents] [Previous Chapter] [Next Chapter] [Main Page] [Top of Page]