Building ‘Universal Service’ in the Early Bell System: The Reciprocal Developmentof Regional Urban Systems and Long Distance Telephone Networks1

David F. WeimanDepartment of Economics, Queens College and the Graduate Center, CUNY

and the Social Science Research Council

To be Submitted to “History Matters: Economic Growth, Technology, and Population: AConference in Honor of Paul A. David,” June 2-3, 2000. Please do not cite or quote without

permission of the author.

1

In 1907 at the dawn of an earlier technological epoch, Theodore Vail pronounced AT&T’s

ambition to build an integrated (or seamless) electronic communications network of continental

proportion and, I should add, under Bell control. Vail’s ambitious goal, what he termed Universal

Service, would require vast investments in building and upgrading long distance lines and

fundamental innovations especially in transmission technology.2 Consequently, it would not be

fully realized until the 1930s. Yet, despite the uneven regional diffusion of telephone service and

regional telephone networks riven by competition, Vail insisted that the Bell System had

“assimilated itself into and in fact become the nervous system” of American business.3

Vail’s claim could be readily dismissed as mere hype, echoing the pretensions of earlier and

later entrepreneurs promoting new electronic communications and information technologies.4

While not unsympathetic to this reading, I believe that it contained an important kernel of truth.

In the interim between his reigns over AT&T, Bell managers and engineers had built a more or

less integrated transregional telephone network – a hierarchy of hubs-and-spokes that paralleled

the maturing US urban system.5 When viewed from this perspective, Vail insightfully grasped the

“wholesale connection” between regional urban systems and telephone networks. The diffusion

of long distance telephone service, it connotes, followed and reinforced established patterns of

wholesale trade and so the spatial organization of economic activity at the time.

In this paper I elaborate this “wholesale connection” through an analytic rendition of the

strategies and insights of key managers and engineers of the newly formed AT&T Company

between 1887 and 1914. The argument unfolds in four parts. The first section reviews the basic

1

technology of early telephone service and Bell’s traffic studies, which suggest an affinity between

long distance telephone service and wholesale trade. Sections two and three explain the

hierarchical organization of regional long distance telephone networks by the “natural” flows of

long distance telephone traffic and ultimately increasing returns to long distance telephone service.

I conclude by specifying the spatial-economic, not technical, sources of increasing returns, which

correspond to the very conditions fostering metropolitan or regional urban system development.

1. The Wholesale Connection in Early Long Distance Telephone Service

The parallel between regional urban systems and long distance telephone networks is not

incidental, a mere correlation. By substituting for direct face-to-face contact, the telephone

carried the very economic transactions vital to metropolitan intermediaries – open-ended

negotiations and non-standardized information. Moreover, because of the novelty and expense of

long distance telephone service during this formative period, it was used for little else.6

In 1887 Edward J. Hall, then general manager of the newly formed AT&T Company, clearly

grasped this “wholesale connection.” Responding to the very pressing question of why customers

would purchase more costly long distance telephone as opposed to telegraph service, Hall

specified its “separate and distinct” demand or market segment.7 “When the nature of the

business requires personal communication, question and answer, the railroad or the telephone line

must be used, and this is our field: quick communication with instantaneous replies and prolonged

personal interviews.” In other words, he observed, “If the long-distance telephone competes with

anything, it is with the railroad,” and not the telegraph. Once “people learn its uses,” he

2

predicted,” long distance telephone service will become “the most important factor in the

transaction of business between distant points.”

As a corollary, Hall envisioned regional long distance networks in terms of a hierarchy of hub-

and-spoke systems, each anchored by a commercial center and bound by the city’s hinterland or

trade area. Writing to the president of AT&T’s parent company American Bell Telephone in

1888, he recommended the construction of two regional networks, one centered around New

York and the other around Chicago. “Included within each of these two great circles,” he

elaborated, “would be a host of smaller ones centering at the various large cities from which the

business of a state or section radiates.” Regional and local toll centers played a pivotal role in his

plan, as they would mediate calls within and between their territory and so forge an integrated

national network, that is “universal service.”

“Universal service” would enable telephone customers, regardless of their location, to

communicate directly through the flow of traffic between interlocking networks, distinguished by

their geographic range, physical design, and mode of service. At the local level, a city or district

of a large metropolitan center, the exchange network connected the telephones of residential and

business customers to a central office, which housed the switching equipment. Toll and long

distance trunk lines, in turn, linked central offices within and between the territories of telephone

operating companies and thereby formed larger geographical networks. Local toll networks

embraced economically and geographically proximate centers, and in turn were joined by toll and

long distance trunk lines, typically between the largest centers, to form regional and transregional

(e.g., national or global) networks.

3

The distinct domains of local exchange and toll and long distance service suggest an obvious

parallel to the differences between retail and wholesale trade. Like trips to the grocer or drug

store, households tended to restrict their telephone calling to the immediate vicinity, often only to

the neighborhood level. As historical studies of its usage have clearly demonstrated, the

telephone did not directly expand households’ geographic horizons, but instead merely cemented

their local ties.8 Moreover, exchange service satisfied the spontaneous demand for local

interactions by completing these connections almost instantaneously – that is, in less than a minute

or while the caller remained on the line.9

The speed of service in the era of manual switching technology – the time necessary to

complete connections – reflected the greater facility of mediating local transactions and, in turn,

the relative simplicity of exchange technology and operating methods. Subscribers gained prompt

access to local operators, because their telephones were directly wired or looped into the central

office switchboard.10 To signal the demand for a connection, the caller just picked up the

telephone receiver and within seconds reached an operator. In most cases the same operator

could complete the entire transaction. Through a multiple switchboard, she could reach the

connecting jack for every exchange subscriber and just bridged the circuit linking the caller to the

desired party.11

As in wholesaling, toll and long distance service mediated transactions over greater distances,

within and between operating company territories respectively. These far-flung interactions were

more complex, often following indirect routes, and so demanded additional steps and time. Even

under the best of circumstances, operators could require almost 10 minutes to make the desired

connection.12 Consequently, when customers contacted an operator to request a toll call, they in

4

effect issued an order to be fulfilled at a later, specified time. Callers, in other words, simply

reserved a time slot on toll trunk lines. In turn, by accumulating these demands, toll centers

coordinated the flow of traffic to and from, as well as within, their territory.

Toll service required greater mediation because of the hierarchical design and operation of the

network. Unlike in an urban exchange, subscribers acquired access to the toll network indirectly

through a limited number of trunk lines connecting exchange and toll switchboards.13 To

complete a toll transaction, the exchange operator funneled the call to a toll operator, who

recorded the desired connection and directed the call along the appropriate route to its final

destination. The speed of service depended on the number of intermediate switches or relays

necessary to reach the end point and the available trunk line circuits at each step along the way.

Excess demand for toll circuits at any juncture caused delays and the formation of queues. Thus,

as explained in the next section, the efficient organization of the toll network operated according

to the principles of pooling bulk transactions analogous to those in wholesale trade.

Despite the potential geographic scope of telephone connections, the vast majority of traffic

before 1930 was local, restricted to the domain of the local exchange. Between 1890 and 1920

toll and long distance calls accounted for two to three percent of all telephone calls in the US (see

Figure 1).14 Their share barely exceeded four percent by 1929 despite the rapid growth of long

distance service in the 1920s. Nonetheless, the utilization of the toll network served as a keen

barometer of aggregate economic activity. The volume of toll calls tracked the seasonal flows

trade and credit, as well as the peaks and troughs of the business cycle. As evidence of the latter,

the relative number of toll and long distance calls declined sharply during cyclical downturns,

especially the Great Depressions of the 1890s and 1930s.15

5

Correlation in this case corresponds to causation because of the selective demand for toll and

long distance telephone service. Local exchange service mediated both business and personal

interactions and so issued mixed signals. By contrast, toll and long distance services were an

“economic instrument,” carrying almost exclusively transactions within and between business

enterprises.16 Even toll and long distance calls made from residences or vacation resorts, while

ostensibly personal, were often for business purposes, as managers used the convenience of the

telephone to keep tabs on their distant operations.

Business demand for toll connections can be ascertained quantitatively from traffic studies of

Bell operating companies, which indicate the frequency of toll calls by type of customer –

residential versus business – and even by “class” of business. An early report from the Bell

exchange in Buffalo provides such a detailed breakdown of outward toll traffic in 1891.

Residential customers, including doctors, accounted for almost a quarter of exchange subscribers

and approximately 10 percent of local calls. Yet, only 17 percent of these households called

beyond the local level, and made on average only two toll calls per month (see Table 1).

Businesses, by contrast, utilized telephone service more intensively, including the toll and long

distance networks. Over half of all business customers placed at least one non-local call a month,

although most firms did not call frequently. The largest users of toll and long distance services

were firms that figured significantly in the nexus of wholesale trade: hotels, telegraph companies,

commission merchants, specialized wholesalers, banks, and shippers (see Table 1). Firms in retail

trade (such as carting, stationary supplies, and dry goods) made fewer toll calls, as they conducted

transactions on a more local (i.e., exchange) level.17

6

The pattern of usage in Table 2, based on a smaller exchange in upstate New York, shows an

even sharper disparity in business and residential demand for local and toll services. Households

accounted for 44 percent of all exchange customers, but only a third used the toll network during

the month. Of those who did, only 5 percent placed more than five toll calls.

Businesses obviously constituted the principal source of demand for toll service. Almost two-

thirds of business subscribers used the toll network, and their calls represented nearly 90 percent

of the toll traffic originating from subscriber telephones (see Table 2). Although these data do not

identify the actual source of toll traffic, they do show the avid demand for long distance

connections by a small segment of the business community. Half of all business users made fewer

than 10 calls per month. At the opposite end of the spectrum, the largest 5 percent made at least

20 toll calls per month, and accounted for 40 percent of toll calls by businesses and over a third of

all toll calls in the district.

Finally, Table 3 lists the largest users of toll and long distance services in New York City in

the mid-1920s and so further specifies some of the principal sources of business demand. With a

few exceptions, two types of enterprises predominate. The first are hotels. Without discounting

their tourist trade, hotels, then as today, furnished business executives and sales agents with

lodging and vital services during their routine trips to the nation’s economic center. The second

set of establishments included the corporate offices, often headquarter facilities, of national and

multi-national firms. In both cases, toll connections were essential to conduct wholesale trade,

whether to reach retail customers in the trade area or to keep close contact with distant

production facilities and distribution centers. Notably missing on the list are intermediaries in

New York’s financial and mercantile districts. As Bell’s head of marketing noted, his tabulation

7

would most likely exclude these firms, because their intensive demand for long distance

connections justified leasing toll lines directly from AT&T.18

2. The Bell Standard: The Design and Operation of Regional Toll Networks

Under Vail’s predecessor Frederick P. Fish, AT&T’s Engineering Department had already

launched a thorough reorganization of Bell’s operating divisions and their long distance networks.

In 1904 it circulated a manual on the proper design and operation of regional toll networks.19

Prepared by Thomas Doolittle head of the Toll Traffic Studies Division, these guidelines would

more seamlessly integrate local toll networks into larger regional and transregional systems and

thereby maintain the uninterrupted flow of traffic between any points.

Like “springs, brooks, and rivers,” Doolittle observed, toll traffic obeyed forces “as immutable

as the laws of nature.” Despite the choice of metaphor, Doolittle acknowledged the fundamentally

spatial-economic nature of the “laws” governing toll traffic and networks. Bell engineers, of

course, adapted their networks to the exigencies of the natural geography – the physical distance

or natural barriers separating centers. Still, as Doolittle insisted, an efficient regional network

“should follow the natural trend of the business within its territory.” In other words, when

designed and operated according to Bell standards, the network would accommodate the flow of

traffic among urban centers and so follow the spatial-economic principles governing their

interactions.

i) Nodality and hierarchy in toll networks

Significantly, the adoption of Bell standards forged a hierarchical order of toll centers

according to their nodality and function in the network. The term nodality connotes the

8

accessibility of a center to other places in a region. Intuitively, it conveys the sense of “being

close,” a general proximity to any destination, not just one or several. This complex relationship

embraces: 1) a metric to measure the distances between centers, 2) an aggregation principle to

translate the matrix of distances into a more or less unitary ranking, and 3) a reference point to

delimit the spatial domain. Depending on the geographic scale, a center might be deemed more or

less nodal. It may be the most accessible point in a local area, but occupy a more peripheral place

in a larger regional context.

In the geography of networks whether communications or transportation, proximity is a

deceptively simple term. If it corresponds to physical distance, then nodality is reduced to the site

characteristics of a center, such as a central location within the territory. In explaining the

efficient routing of traffic, Doolittle seemed to embrace this definition. The “shortest and best

route” for directing traffic, he maintained, minimized the “circuit mileage” or physical distance

traveled. In explicating this rule, however, he subtly shifted tone and replaced this purely physical

notion of distance with one rooted in the very structure of the network. An efficient path, he

continued, required the “least number of [intermediate] switches,” and in the case of a tie, the

“best route” was furnished “with the greater number of circuits.”

Proximity in the latter sense is synonymous with connectivity, the “complications and cost” of

linking two points in the network, and depends on the number direct pathways or circuits joining

them.20 By avoiding relays at intermediate points, direct routes minimized the time to complete

connections, the likelihood of errors, and the risk of congestion delays along the way. Also,

fewer switches diminished the attenuation and distortion of the voice signal and so enhanced the

clarity of the message.21 Finally, stringing additional circuits along toll lines to a destination

9

significantly lowered bottlenecks at toll centers, while direct lines to other nodes in the network

furnished more efficient secondary routes in case of congestion or service interruptions on

primary ones.

Thus, a nodal center insured immediate contact to any destination in the network, because it

furnished more direct connections to a wider range of points, especially other pivotal nodes. The

case study analyzed by Doolittle neatly illustrates this principle. The circuit map in Figure 2

depicts a regional network, redesigned by Doolittle according to Bell standards.22 It contains two

kinds of centers. Large urban exchanges (denoted by lettered boxes) serve as toll centers

mediating the flow of traffic through the network. Smaller centers are furnished with branch

exchanges (numbered boxes and unfilled circles) or toll telephone stations (filled dots). The latter

were nothing more than public telephones, centrally located at a general or drug store, where

residents could make and receive long distance calls. Finally, the lines represent wire circuits, the

physical medium for transmitting messages between centers.

Comparing the site characteristics and connectivity of exchanges in Figure 2 sharply contrasts

physical and spatial-economic notions of distance and nodality. In terms of their geographic

location, exchanges I and J are located closer to the center of the territory, while A and D are

situated on the perimeter. The latter two, however, are more nodal, when ranked by the facility

of communication – the “cost and complications” of completing connections – such as the

average number of switches or routes connecting a center to any point in the territory (see column

2 in panel A of Table 4). Following the primary or most efficient routes as designated by

Doolittle, traffic to or from centers A and D required on average less than one intermediate switch

10

(0.83 and 0.70 respectively). By contrast, routes connecting J and I to other points in the

network were more roundabout, averaging 1.04 and 1.43 intermediate switches.

This index abstracts from the quality of toll connections in relation to the spatial patterns of

traffic and so understates differences in nodality between these places (see panel B of Table 4).

Exchanges A and D were furnished with direct circuits to almost twice as many destinations in the

network as were J and I (35 and 40 percent of all centers versus 22 and 17 percent), and the bulk

of their traffic (between 86 and 90 percent of total calls) flowed along these uninterrupted

routes.23 Moreover, multiple circuits along the most heavily traveled lines (such as from A to U

and B and from D to B and H) greatly enhanced transmission capacity and hence the reliability

and speed of service. Finally, with one exception, the more nodal centers afforded a greater range

of auxiliary routes (see columns 3 and 4 in panel A of Table 4).24 The anomaly, exchange J,

illustrates a potential benefit of its central location at the cross-roads of routes connecting other

centers.

At the opposite end of the spectrum, centers W and K were marginal because of their spatial

isolation, the sparsity of their network connections. Situated on the very edge of the network,

exchange W gained access to other centers only through the line joining it to exchange A. This

indirect link required an additional relay to complete calls to other centers and accordingly

diminished its nodality as measured by the average number of intermediate switches (from 0.8 for

A to 1.8 for W). Significantly, over 40 percent of the traffic to and from W required at least one

intermediate switch, and 7 percent, at least two.

Despite an additional outlet, exchange K was, if anything, more marginal. In fact, Doolittle

had replaced the direct circuit from K to V with an indirect connection, and so restricted its

11

access to a single circuit each on the other two lines. Although the bulk of traffic to and from K

(75 percent) was direct, requiring no intermediate switching, 16 percent of all calls had to be

relayed at least two times. Another consequence of its peripheral position was the limited range

of tertiary paths (to only 4 percent of other points) and so the greater risk of isolation in case of

damage to primary and secondary lines.

The design and organization of the network also illustrates the dual relationship between the

nodality and hierarchy of centers, the latter referring to functional position of toll centers in

mediating the flow of traffic. Nodal centers assumed more pivotal roles in controlling and

coordinating traffic through the network (e.g., the routing of toll calls and financial accounting).

Conversely, marginal centers reached other points indirectly via more nodal exchanges, and so

were rendered, in effect, tributary to them. The nodality of centers, in turn, defined their spatial

range, the scope of network traffic which they mediated or equivalently the number and location

of places in their tributary area.

The hierarchy of centers operated immediately in the formation of exchange districts, which

essentially extended the domain of large urban exchanges to encompass smaller nearby towns and

villages (as illustrated by the irregular boundaries surrounding toll centers in Figure 2).

Implementing AT&T’s center checking system, Doolittle removed branch exchanges and toll

stations from the direct circuits connecting toll centers and placed them on “tributary” lines wired

into switchboards at an adjacent exchange.25 The very design of the network treated toll stations

“practically as subscribers of the exchange centre” in terms of their access to toll lines and thereby

increased the financial and operational control of toll centers over traffic in their immediate

vicinity.

12

Network design imposed a similar hierarchy among toll centers. The toll center in exchange

A, as noted above, mediated the traffic of nearby exchange W. Similarly, by removing the direct

circuit between Q and F, Doolittle channeled the traffic of the former exchange through the center

at point D. Such relationships, however, were less rigid than those defined by the center checking

system. For example, with the completion of the proposed circuit between centers E and B, the

former exchange would gain comparable access to the network through two nodes, not one.

Routing instructions, which directed traffic through intermediate points to its final destination,

reflected these more flexible hierarchical relationships. Following Doolittle’s “simple” rule, they

specified the most efficient paths between all points in the network. When confronted with

alternatives, this early application of the turnpike principle tended to channel traffic through more

nodal toll centers, even if the resulting routes did not minimize the physical distance traveled. In

other words, greater connectivity often trumped mere proximity in governing the flow of traffic,

and in turn greatly expanded the spatial range of more nodal centers. To cite an example,

Doolittle routed calls between points D and U through center A. Despite some backtracking, this

path required only one intermediate switch and offered a greater number of available circuits. In

the reverse direction, traffic between points A and H was routed through J, although the

alternative through D was nearly as efficient.

Following the “evolution and systematic development” of toll operating methods, especially

for handling calls between geographically proximate centers, traffic engineers at New England

Telephone elaborated Doolittle’s guidelines.26 To exploit fully these potential economies, their

“toll center plan” relayed the traffic of smaller centers through toll offices at nearby larger ones.

This method substantially lowered circuit (or capital) and labor requirements and so total unit

13

costs. Moreover, it extended the hierarchical organization of toll networks to encompass the

relationships among toll centers. Applied to the territory of New England Telephone, it

consolidated control and coordination functions in its 79 largest toll centers, as compared to

nearly one center for each of its 730 exchanges previously. The resulting design clearly

anticipated the top-down structure of Bell’s General Switching Plan introduced in the late 1920s.

ii) Proximate sources of nodality in long distance telephone networks

Following his most fundamental tenet, Doolittle organized the network in accordance with the

“trend of traffic.” The implied relationship between the nodality of center, its hierarchical order,

and the flow of traffic that it mediated reflects the influence of increasing returns, due in part to

large indivisible investments in transmission and switching facilities. For example, direct access to

the network via a toll center and lines to other centers entailed substantial capital outlays: laying

poles, stringing higher quality wire circuits, and installing or upgrading central office equipment.

Profitability, therefore, depended on capacity utilization and hence traffic density. At a minimum,

the volume of traffic was expected to cover fixed and quasi-fixed costs, including depreciation

and maintenance of the plant and employment of full-time operators and supervisory personnel.

If markets were too small to support an exchange let alone additional toll center facilities, they

could be furnished with toll stations, wired into an adjacent exchange by lower quality circuits.

Acquiring even this limited access depended on whether projected revenues exceeded break-even

levels, determined by the overhead costs of the physical plant. Doolittle did admit one exception,

which seemingly allowed an independent role for geography. “It is sometimes advisable,” he

counseled, “to build lines that do not promise an immediate return, in order to round out a

system.”27 Although he did not elaborate this criterion, it suggests extending the network to

14

nearby towns, regardless of the expected volume or trend of traffic. At the very least, the

incremental cost of building these lines would be small because of the short distances involved.

Yet, as shown in the concluding section, he implicitly offered an alternative interpretation,

consistent with his principle relating network design to the spatial-economic patterns of toll

traffic.

The center checking system, which delineated the geographic boundaries of exchange

districts, vividly illustrates the imprint of traffic patterns on network design. “Toll stations,”

Doolittle explained, “should be connected to the exchange center to which their traffic is tributary,

but with due consideration to the length of the ... circuit.”28 Despite the latter qualification, these

connections represented the net flow of traffic between toll stations and adjacent centers. The

charts in Figure 3 illustrate his method for distributing the stations located between centers A and

B. For each station, the horizontal lines and corresponding magnitudes indicate the volume of

traffic to and from the connecting points (e.g., 3 calls between stations A-27 and -23). Although

not necessarily destined for point A, each station’s traffic flowed overwhelmingly in that direction.

By accommodating this trend, the proposed tributary circuits minimized the cost of making

connections, i.e., the number of intermediate switches required and not simply the physical

distance traveled (compare the initial and proposed position of station A-27 in the circuit

diagrams in the lower half of the Figure 3).

The same reasoning governed the construction of direct circuits between toll centers and so

determined their accessibility to all points in the network. To illustrate, Doolittle proposed

connecting exchanges A and D by a direct line, bypassing intermediate exchanges along the way.

The volume of traffic between A and D averaged 18 calls per day. Assuming normal demand

15

growth over the planning horizon (which amounted to adding 60 percent to the actual load), this

single circuit eventually would support approximately 29 calls daily, just under its rated capacity.

Conversely, the marginality of a center followed from its limited interactions with other

centers. In one instance Doolittle removed the direct circuit between exchanges K and V, because

the line carried a low load, only 16 calls per day, and 75 percent of the calls originated at point U,

not K.29 Just like small centers furnished only with toll stations, he explained, center K did not

warrant a direct line for traffic in this direction, but would operate under the control of the more

dominant adjacent center U.

Doolittle’s method does, however, contain one potential hitch. The addition (or removal) of a

direct circuit between two centers depended on the total flow of traffic between them, equal to

their “own” traffic – calls between parties within each district – and relay traffic destined for other

points. The inclusion of the latter element implies that the nodality of a center could simply derive

from its designation as a switching point for traffic to and from other places. If the system of

routing instructions, in fact, governs the “trend of traffic,” then the very notion of nodality is

either rendered tautological or is made contingent on an exogenous factor, such as geographic

centrality. In other words, the network would be ordered by an external principle and not by the

endogenous interactions among centers, the “natural laws” to which Doolittle had alluded.

A closer inspection of his method for designing the network resolves the dilemma. Instead of

circular reasoning, his procedures imply a self-reinforcing mechanism that gives priority to a

center’s own traffic. In every instance, this component – a center’s total number of inward and

outward calls – accounted for the majority of its traffic with other points in the network and

therefore made the largest contribution to achieving the necessary demand thresholds. For

16

example, in the case of the proposed direct circuit between centers A and D, over half of the calls

(60 percent) represented their “own” or direct traffic, and the rest consisted of relay traffic to

outlying points in both directions.30

Moreover, Doolittle determined the system of routing instructions and thus the volume of

relay traffic between centers through a sequential process, which followed and hence bolstered the

spatial patterns of own traffic. He initially furnished centers with a sufficient number of direct

circuits to handle their own traffic. This preliminary network established the incipient nodality

and hierarchy of centers. Applying the principles of routing traffic, then, he determined the most

efficient paths for relaying calls through the network. Consequently, the flow of relay traffic

favored routes through more dominant centers, determined by their connectivity and hence the

scale and scope of their direct calls. Thus, relay traffic tended to augment, not offset, the flow of

own traffic and to enhance the nodality of centers based on the spatial patterns of their direct

interactions with other centers.

The one exception, as is often the case, helps to prove the rule. The flow of traffic from

center T to point D and beyond was routed through exchange H and not F, even though the latter

pathway potentially offered more circuits and fewer intermediate switches (see Figure 2). This

decision was based on the larger volume of relay traffic through H, which also included calls to

and from points in the other direction, and hence on its more central location. Geography played

a decisive role in this case, precisely because centers F and H supported approximately the same

volume of traffic (108.4 and 100.2 calls daily) and so did not dominate each other along this

critical dimension. Consequently, Doolittle reasoned, H would probably be furnished with

17

additional switching capacity to handle its other relay traffic, and so could more easily

accommodate the rest.

Thus, Doolittle’s analysis specified, at least implicitly, the “trend of traffic” by the matrix of

inward and outward calls for each center and so by factors independent of routing instructions. It

yields, moreover, an alternative representation of the nodality and hierarchy of centers according

to their own traffic. The nodal position of centers A and D, for example, ultimately derived from

the larger volume and greater diversity of their inward and outward traffic (see the first and last

columns in panel A of Table 4). These centers supported 200 to 300 calls daily, and the four

largest routes accounted for only 60 percent of the total flow. By contrast, the marginal position

of exchanges W and K accorded with their narrow niche in the network. Both exchanges

supported a significantly lower volume of traffic, only 82 and 34 calls daily. Equally important,

the bulk of their traffic was confined to neighboring centers. For exchange K, over 70 percent of

its traffic flowed back and forth between two adjacent exchanges. Similarly, four nearby centers

accounted for almost all of the traffic to and from W.

This simple case vividly illustrates Jacob Price’s thesis, explaining the formation of entrepots

or metropolises by the “complexity of trade.”31 In fact, his notion corresponds precisely to the

source of nodality in toll networks, for, as shown in the previous section, the size and scope of a

center’s own traffic stemmed from and so mirrored the wholesale trade of its district. Just like in

the Chesapeake colonies, the sheer volume of toll calls was not a sufficient condition to elevate

centers to nodal points, controlling and coordinating traffic through the network. A large number

of calls, restricted to only one or two routes, relegated centers to a tributary position, because

their simple, bilateral transactions could be conducted at distance. A pivotal node or higher order

18

center, just as the term connotes, required more generalized contacts; its interactions, in other

words, could not be so readily confined.

3. Increasing Returns in Toll Networks

The hierarchical organization of regional telephone networks, as suggested above, ultimately

depended on the sources of increasing returns, which created and reinforced the initial advantage

of larger, more dominant centers. Hall, for one, clearly perceived these systemic processes and

their influence. “Developing fully a good system of terminal and branch feeding lines,” he

observed, “... will not only pay in themselves but will add to the business of the main trunk lines,

and every gain in business requiring additional trunk wires means a large gain in profit.”32

Increasing returns, he implies, derived from economies of scale and interdependent demands.

Scale-dependent processes yielded greater efficiencies (the “gain[s] in profit”) from concentrating

traffic through larger toll centers and the trunk lines connecting them. At the same, extending the

network to encompass interdependent users, those bound by a strong “community of interest,”

enlarged the demand for toll services, especially the utilization of trunk lines.33 A properly

designed network would tap these complementary demands and generate substantial externalities,

the unintended or synergistic benefits from their mutual interaction.

Bell engineers tended to regard these forces mechanistically, that is simply in terms of

technology or other intrinsic properties of the network. Yet, through insight or experience, key

officials comprehended the necessary market conditions to realize fully the telephone’s economic

potential. Harnessing these market forces influenced the design of all facets of the network –

plant technology, operating methods, and spatial structure. Thus, when properly conceived, the

19

processes underlying the development of the toll network are more aptly termed “eco-technic”

and exemplify the opposing, but mutually reinforcing tendencies of centralization and dispersion

that propel urban systems and regional development.34

i) Economies of scale

Building toll lines, as noted above, required large, indivisible investments in transmission

capacity. Telephone companies could economize on these outlays in numerous ways, by reducing

the number of circuits on a line and increasing the spacing between poles.35 Still, even the most

rudimentary connections imposed truly fixed costs on operating budgets, such as leasing and

maintaining rights of way. In turn, the growing volume of traffic increased capacity utilization

and lowered unit costs.

On more heavily traveled routes, additional investments in transmission capacity, including the

adoption of new technologies, yielded even greater economies and improvements in service

quality.36 Stringing more miles of wire on existing pole lines, for example, spread fixed costs over

a larger number of circuits and so reduced the unit overhead charges on each. Moreover, through

practice and theoretical analysis, AT&T engineers discovered that larger circuit groups – trunk

lines furnished with multiple circuits – yielded genuine economies of scale. Doolittle’s 1904

bulletin indicated the actual magnitude of these efficiency gains.37 Under “ordinary conditions,” a

line equipped with a single circuit could carry an average load of 30 calls per day. The same line,

furnished with five circuits, could transmit 200 to 225 calls daily or an average of 40 to 45 calls

per circuit. In other words, enlarging the circuit group by a factor of five enhanced the effective

capacity or load of each circuit by 33 to 50 percent.

20

Around the same time, Bell engineers embarked on a program of theoretical research, which

explained the source of and limits to these efficiencies. The reasoning is essentially probabilistic,

an instance of the law of large numbers. Under “normal” conditions, pooling diverse sources of

traffic at a single large center and along larger circuit groups would smooth out random variations

in the timing and duration of individual calls.38 Consequently, plant managers could predict more

accurately average loads over time (e.g., a day, week, and year) and plan network capacity

accordingly. In turn, because demands would exceed plant capacity on an incidental or irregular

basis, individual circuits could carry, on average, higher daily loads without increasing the risk of

congestion and so reducing service quality.39

The adoption of loading coils also generated substantial savings in fixed costs. By diminishing

the attenuation of the voice signal, loading coils economized on the use of copper wire and at the

same time greatly enhanced the quality and range of voice transmission. Additionally, it allowed

the introduction of “phantom” circuits, which created additional transmission capacity “without

any expenditure for new wires.”40 Bell engineers eventually succeeded in forming a single

phantom circuit for every two loaded wire circuits and thus increased the effective capacity of the

latter by fifty per cent. The considerable demand and technological thresholds of this technology,

however, limited its diffusion to Bell’s principal routes, such as between New York and

Chicago.41

Concentrating traffic through larger toll centers also initiated a Smithian dynamic of

technological and organizational change. Denser, more differentiated traffic patterns warranted

the construction of switching and transmission facilities tapered to distinct market segments. In

turn, Bell engineers refined operating methods that exploited more fully the economic potential of

21

these specialized facilities. Despite the greater outlays of fixed capital and employment of

overhead personnel, these innovations magnified the productivity of switchboard operators and

the rate of “throughput” of the toll plant by standardizing and speeding up manual operations.42

The centralization of more specialized equipment and personnel also enabled Bell companies to

offer new services, extending the economic scope of its fixed plant and overhead staff.

An important instance was the evolution of toll operating methods within larger metropolitan

centers. The greater volume of traffic between these nodal points justified the construction of

direct circuits, which greatly enhanced the speed of service and circuit loads. In turn, on denser

routes furnished with multiple circuits, Bell engineers exploited the greater transmission capacity

to improve operating methods. An 1898 memorandum from Doolittle to American Bell President

Hudson described such innovations, the “recording” and “call wire” systems developed by Joseph

Carty of New York Telephone.43 In the former specialized toll operators divided the tasks of

recording information on and overseeing the desired connection. “Recording” operators relieved

local operators of clerical tasks and insured a more thorough accounting of outgoing toll calls.

“Line” operators, responsible for a limited number of trunk lines and destinations, completed

connections more quickly and monitored them more carefully.

The latter innovation dedicated a circuit, the “call wire,” for transactions between operators.

The “call” circuit placed operators in “instant communication” and so increased their productivity

and the average speed of service. Also, by reserving toll circuits exclusively for conversations

between paying customers, it substantially increased their carrying capacity and average daily

loads. Despite these gains in operator and circuit efficiency, the “call wire” system was deployed

only on the routes between Manhattan and Newark and Yonkers, precisely because it devoted a

22

wire circuit on each line to signaling.44 For other toll calls within the metropolitan area, operators

employed the ordinary “ring down” method and communicated with each over toll circuits. As

Carty explained, “while this is not such quick service, it is as quick as can be expected where

trunks are not sufficiently numerous.”45

Bell engineers further refined these procedures to accommodate the markets for toll traffic

within and between metropolitan regions.46 For intra-regional traffic ranging from 50 to 100

kilometers, the “two-number” system approximated the switching methods used in urban

exchanges. Toll calls were routed through a separate office (or switchboard) and network of

trunk lines. Like exchange service, “two-number” operators could complete connections to any

destination in the territory and executed all of the steps: selecting a free trunk line, transmitting

the called number to the receiving operator, and accounting for the transaction. Moreover,

connections were made instantaneously, that is while customers remained on the line.

The savings in manual operations and time depended largely on the formation of specialized

facilities, not technical innovations. An essential ingredient was the publication of a directory,

which listed the telephone number of subscribers by central office (or location). The telephone

directory enabled customers to request connections without the assistance of information

operators and greatly simplified the task of receiving operators. These conditions, in turn,

restricted its use to metropolitan trade areas.47 The larger, more regular flow of intra-regional

traffic justified the investment in fixed capital and in compiling a directory of frequently called

numbers. On the flip side, the greater density of the metropolitan market offered firms an

inducement to purchase a listing and so to advertise in the telephone company’s classified

directory.

23

In a similar fashion, the “single-ticket” method rationalized the routing of long-haul traffic

between metropolitan centers. Operators at the point of origin assumed full control over

recording and routing toll connections. They prepared the tickets for billing customers and issued

routing instructions to receiving operators at intermediate and destination points. The latter

simply executed manual operations delivered over the long distance circuit. The greater

centralization of control substantially reduced the time and cost required to make long distance

connections, as well as potential operator error. More rapid, reliable service translated into higher

levels of capacity utilization, as operators required less time for signaling.

Like other innovations, the diffusion of this method followed the consolidation of AT&T’s

national network and proceeded gradually down the hierarchy of toll centers and trunk lines. It

was initially adopted on the denser routes between metropolitan centers, which were equipped

with high gauge wires and loading coils. These technologies extended the geographic range of

long distance transmission and so allowed line operators to direct traffic over longer distances.

Also, effective communication among operators demanded uniform technologies and operating

methods. The adoption of Bell standards, however, depended on the thorough integration of

operating companies into the national network and hence on the volume of their interregional

traffic.

ii) Interdependent demands and network externalities

Alongside the technological and organizational processes fostering centralization, Bell officials

and engineers also recognized an opposing tendency to extend the network geographically.

Characteristic of other communications and transport networks, this centrifugal force derived

from the interdependent demands for telephone service. As Doolittle repeatedly claimed, a vaster

24

network, connecting more people over a wider area, would increase the potential use of the

telephone and hence its value to customers. This abstract formulation, simply relating individual

demands for telephone service to the number of potential users and connections, ignores the

actual spatial-economic patterns of interaction that determine the distribution of benefits from a

larger network. From his vantage point in AT&T’s Toll Traffic Studies Division, Doolittle would

elaborate Hall’s original insight and conceive this principle in terms of the complementary

demands for toll service within urban regions.

Interdependent demands, Doolittle explained, depended on the geographic range of economic

and social relationships. Within urban centers, the highly localized nature of social interactions

tended to restrict the systemic benefits of widening telephone networks. Residential customers

called frequently within their immediate vicinity, but except for occasional transactions requiring

connections to business districts, they rarely contacted subscribers in other neighborhoods. By

way of explanation, Doolittle offered a quaint analogy to the impact of urban annexations: “the

individual use of the telephone is quite similar to the individual use of the streets; each [person]

has his particular route and rarely deviates from it and would not if the city were many times

greater.”48 Thus, enlarging exchange networks to incorporate new districts would add

subscribers, but not notably increase the demand for service by existing ones.

The toll network, by contrast, mediated more far-flung relationships and therefore was

governed by an opposing principle. After a decade’s experience, he wrote to Joseph Davis, head

of the Engineering Department: “Our records show that the larger the number of places

connected, the larger will be the percentage of people interested in the toll lines.”49 He proceeded

to depict these interdependent demands in terms of a schematic model, which showed the mutual

25

interaction between the short- and long-haul traffic of commercial centers (see Figure 4). In the

top panel the network, consisting only of a trunk line connecting two large centers, is hardly

utilized. In subsequent panels, the construction of toll lines to hinterland cities and towns induces

an increased volume of intra- but especially interregional traffic. The latter flows along the trunk

lines between the large centers and thereby generates scale economies.

Doolittle intended his simple model to promote local network development by Bell operating

companies. Ignoring the systemic benefits from a more extensive network, they often devoted too

few resources to building complementary toll lines in their territory. Even if tributary lines were

not profitable when reckoned on a stand-alone basis, he showed, they would induce greater

utilization of the network, especially trunk lines, and so enhance aggregate earnings. In effect, he

advocated subsidizing the construction of tributary connections to smaller centers with the greater

earnings on trunk lines between larger cities.50

Moreover, as his model implies, these internal transfers of funds are not truly subsidies.

Rather, they reflect the spatial separation between the source of the network externality –

investments in tributary connections – and its realization – greater profits on more heavily utilized

trunk lines. In the absence of a rate structure based on value of service, accounting for and thus

realizing these systemic benefits would demand a more extensive, vertically integrated network.

In turn, this vital relationship specified the integral components of areal networks and so

delineated their spatial boundaries systemically, that is in terms of the geographic scope of these

reciprocal demands.51

An ambitious program of network development, proposed by New England Telephone,

exemplifies Doolittle’s model and concretely demonstrates the potential benefits from expanded

26

toll connections.52 The company would furnish smaller cities and towns with rudimentary

exchanges, offering limited local service. These centers, it was acknowledged, would generate

few outgoing calls, the basis upon which Bell companies charged customers and so reckoned the

revenues and net earnings of toll lines. They would, however, attract a substantial volume of

incoming traffic from larger commercial centers and thereby increase the overall utilization of the

network and aggregate revenues. To determine prospective sites, then, Doolittle projected their

contribution to toll revenues in both directions.

Bell’s programs for marketing toll services elaborate the interdependent demands of business

users for toll services. Although prone to overstatement, they depict innovative ways, in which

firms in the mercantile sector could employ telephone and related Bell services to lower their

costs and expand the economic and geographic scope of their trade. Rather than more abstract

characterizations, these concrete examples demonstrate why certain firms would attach greater

value to and so use more frequently, a larger, more extensive network.

The Key Town Sales Plan, in the words of an AT&T commercial engineer, was “a systematic

plan for handling business ... based on the experience of those who were actually doing the

selling.”53 Instead of simply assisting sales agents in arranging business trips, the telephone played

a more integral role in their annual routine. At the principal center of each territory, sales agents

reached customers in tributary towns more easily and inexpensively through the toll network.

Relying on telephone contact for routine transactions such as collecting orders, they reserved

occasional, longer business trips for more complex deals, negotiating price schedules or the

introduction of new products. Using this method, sales agents “covered larger territories and did

more business,” but their “costs in relation to sales were extremely low.”54

27

To assist sales agents, AT&T and its operating companies prepared maps of trade territories

and their key towns or “calling points” that followed the contours of the regional urban system

(see the sample map in Figure 5). Additionally, Bell designed programs that facilitated these

transactions. To reduce delays in securing connections and to routinize their planning, sales

agents could submit, in advance, sequenced lists of periodically called numbers. Before

embarking on their trip, agents would reserve space at toll centers by notifying the chief operator

when they would arrive. Bell also offered a credit plan that billed toll charges to the sales office.

Finally, its classified business directories furnished sales agents with vital information on potential

customers and new market areas.

Bell also promoted toll service to rural merchants, who occupied the lowest rung in the

mercantile hierarchy. Its agents, for example, responded creatively to the extension of rural postal

service, which brought mail order companies into closer contact with farm households and so

threatened the retail trade of country storeowners. To assist their customers, they devised a

program combining the toll network and the postal delivery system to furnish the equivalent of

fast freight service to farm households.55 Through telephone calls, rural merchants could maintain

more regular contact with their customers, and use “carriers on [rural] routes as expressman ...

[to] guarantee quick delivery.”

In the other direction, toll service expedited transactions between rural merchants and their

wholesale suppliers. By ordering over the telephone, rural merchants could restock their shelves

in response to consumers’ demands and so economize on their inventory holdings. This system of

“hand-to-mouth” purchasing availed retailers of the larger, more diverse stores of wholesalers,

while it enabled the latter to exploit economies of bulk purchasing and distribution.

28

4. Market Sources of Increasing Returns

Like the systemic processes of urban systems development, increasing returns in the telephone

network were the product of economies of coordination and specialization. The former

correspond to the benefits from pooling traffic at nodal centers in the network, and the latter, to

the agglomeration effects of vertical disintegration and new market development. Specifying

these market processes more precisely reinforces and elaborates the conclusion of section two.

The nodality and hierarchy of centers depended on the “quality of economic activity” in their trade

area and resulting greater “complexity of trade,” and not simply on the volume of exports of a

single staple product.56

Through experience, traffic engineers not only detected scale economies on denser trunk lines,

but also specified conditions necessary to realize them. In particular, they found that average

costs or conversely normal circuit capacity depended critically on seasonal fluctuations in

demands or loads. As Frank Fowle, a regular contributor to Telephony, observed, for a given

grade of service (e.g., speed in completing connections and risk of congestion delays), capacity

utilization varied inversely with the load factor – the ratio of daily average-to-peak demands.57

Other things being equal, “an uneven load curve with prominent peaks and large, abrupt changes

in demand” would either reduce the average speed of service or the unit capacity of circuits.

The assumptions underlying the theoretical analysis of traffic flows, referred to above, offered

a sharper formulation of these conditions. Larger circuit groups yielded the predicted economies,

when the probability of congestion delays was small and when demands during the relevant

interval were uncorrelated with each other and so could be regarded as essentially “random”

events.58 The latter is best illustrated by the opposite case of perfectly synchronized calling

29

patterns. If the same event triggered large numbers of users to place toll calls simultaneously,

then centers would be flooded with demand, and congestion delays would mount. Such extreme

events are uncommon, occurring usually during a holiday or in the aftermath of a disaster.

Yet, under less extraordinary circumstances, calls could still exhibit high degrees of

correlation over time because of the seasonality of economic activity. For example, in a

specialized agricultural region, the daily volume of traffic multiplied during the harvest season, as

farmers and intermediaries used toll lines to negotiate the sale and movement of crops. The

resulting peak loads during busy periods of the day (typically mid-morning and early afternoons)

severely strained the operation of toll centers.59 To minimize congestion delays at these busy

times, centers would mostly likely be burdened with excessive capacity during the rest of the year

and charge higher rates. Alternatively, furnished with fewer circuits, they would experience large

excess demands at peak periods, which violated both conditions. Similar strains would arise in

regions dominated by a few, very large users, such as a company town, where demand patterns

reflected a single firm’s specific production schedules or the seasonality of its purchases and

sales.60

Realizing the benefits of pooling traffic, therefore, depended on the very structure of demand

for toll service and hence of the regional economy. Demands approximated the theoretical ideal

in more densely developed regions embracing specialized, interdependent economic sectors.

Under these spatial economic conditions, no single industry or firm would thoroughly dominate

the use of toll lines and so impose its seasonal demands on aggregate levels. Moreover, the

presence of many, varied sources of demand, each following distinctive but complementary

seasonal patterns, would insure a smoother distribution of calls over time.61

30

Obviously, no area completely avoided seasonal influences; however, the resulting overflow

traffic was less disruptive where centers were furnished with numerous, efficient alternate routes.

The availability of such secondary paths, as noted above, depended on both the scale and

geographic scope of a center’s own traffic, the diverse sources and destinations of its inward and

outward calls. This condition, in turn, was satisfied only in more developed, diversified regional

economies and derived ultimately from the backward, forward, and consumption linkages of the

region’s economic base.

In other words, the realization of these interdependent demands and hence of increasing

returns depended on the extent of metropolitan or urban systems development. The planned

extension of the network by New England Telephone, mentioned above, clearly hinged on the

dense volume of trade between metropolitan and hinterland centers it its territory. Unlike in

Doolittle’s schematic model, however, the company expected trade to flow from large to small

centers along feeder lines, and not vice versa. This trend suggests that users in the former, not the

latter, valued more highly the greater range of toll connections and, in turn, more strongly

advocated the extension of the network.

A later report by AT&T engineers stressed this very point. “Unless ... complete facilities [are]

provided in towns of all kinds and characteristics, the value of telephone service in places of

higher commercial activity will be materially lessened because of the absence of exchanges in

towns of smaller activity.”62 Significantly, the predicted asymmetric flow of traffic from centers of

“higher” to “lower” commercial activity correspond to the spatial range of their mercantile

functions and hierarchical order in the urban system.63

31

Doolittle’s very analysis of interdependent demands depended on these spatial-economic

relationships, even though his own formulations often obscured the connection. In his analogy to

pedestrian traffic, for example, he incorrectly attributed interdependent demands to a particular

geographic network. In fact, it only showed how the use of the telephone by residential

subscribers paralleled the highly circumscribed fields of personal economic and social interactions.

By contrast, business customers, especially intermediaries in larger commercial centers, demanded

the most extensive range of telephone connections both within and beyond city limits. In local

markets with dual service these “core” users typically subscribed to the competing telephone

companies to gain access to all points. Additionally, to obtain toll connections to nearby cities

and towns, they willingly accepted hefty rate increases for extended Bell service, and, if necessary,

even committed resources to establish their own independent companies.64

In other words, interdependent demands characterized both exchange and toll networks, but

were more visible in the latter, whose use was dominated by “core” business customers. Through

the telephone, wholesale merchants and other intermediaries could maintain closer contact with

clients in their trade area. Substituting more frequent telephone calls for costly, time-consuming

business trips, they could offer more prompt, responsive service. Additionally, by lowering

transactions costs and amplifying the channels along which information flowed, telephone service

potentially deepened and widened trade areas and so enabled merchants in one center to encroach

on the markets of another. Consequently, the telephone became a potent weapon in recurrent

rivalries between urban centers.65

Similarly, Doolittle often insisted on a simple, almost linear relationship between city size and

effective demand for toll service, despite his own evidence which showed a weak, empirical

32

regularity at best.66 In a different context, Bell engineers vehemently denied this correlation, and

instead focused on the “the class and general prosperity of the population” and “business

conditions” as more fundamental determinants of demand.67 Like urban geographers, it seems,

Doolittle used population as a shorthand to gauge the economic functions of a center and in turn

its potential economic interactions with other centers. Significantly, in his comprehensive reviews

of territorial networks, he consulted “Dun’s Reference Book,” and not simply the census of

population, to identify “the most promising places” to extend the network.68 From this source, he

gleaned information on their economic structure: the number, size, and type of banks; “rated”

manufacturing and mercantile establishments; and the size distribution of firms by net worth.

These data identified the likely sources of demand for toll service in each city, as well as the

likely recipients of toll calls from other centers. Unlike population aggregates, they measured

more accurately a center’s total contribution to the growth of the regional toll market. From this

perspective, Doolittle’s criterion of building lines “to round out a system” takes on a rather

different meaning. It implies completing the nexus of economic relationships within urban

regions, and not simply the physical connections between proximate centers. Like Hall, he

recognized that expanding linkages between metropolitan centers and the cities and towns in their

hinterland would enhance traffic flows through the entire network and not only on adjacent lines,

and so generate substantial externalities.69

To reinforce the point, Bell’s marketing strategies for long distance service depended on

complementary transport networks and business services, and not just efficient toll facilities. Key

towns, for example, were chosen because of their “railroad facilities, hotel accommodations,

etc.”70 Their location at nodal points in the rail network facilitated scheduling trips and, if

33

necessary, arranging the shipment and storage of orders. Similarly, hotels offered business

travelers lodging and even auxiliary telephone services. Likewise, the anticipated efficiencies in

rural merchandising and hence the value of toll connections to rural merchants and their suppliers

critically depended on frequent, reliable transport services to insure “just-in-time” deliveries.71

Merchants, obviously, derived little additional benefit from ordering over the telephone, if the

resulting deliveries were delayed because of inefficient or infrequent transport services.

Thus, to engage the systemic processes that fueled increasing returns, long distance telephone

networks required a suitable economic environment, one that embraced a wide range of

specialized, but complementary economic activities and in particular related wholesale, transport,

and communications networks. The economic interdependencies among these spatially proximate

sectors render in a concrete and recognizable form the more abstract notion of a “community of

interest,” underlying the reciprocal demands for toll service. A highly developed regional spatial

division of labor mediated by interlocking networks of transport and communications engendered

the complex seasonal and spatial patterns of transactions that furthered the cumulative

development of regional long distance networks. In turn, the resulting hierarchical organization

of regional telephone networks followed and reinforced metropolitan or urban system

development.

34

1. For helpful suggestions and discussions, I would like to thank Bill Brainard, David Gabel,Stanley Engerman, Albert Hirschman, Daniel Koditschek, Sumner La Croix, Margaret Levenstein,Richard Levin, Kenneth Lipartito, David Pearce, Jean-Laurent Rosenthal, Matthew Shapiro, RossThomson and participants of the Columbia University and University of Michigan EconomicHistory Seminars. Funding from the Yale University Social Science Research Fund and theNational Endowment of the Humanities through fellowship from the Institute for Advance Studyis gratefully acknowledged.

2. In 1907 Vail emphasized the “universality” of service furnished by the Bell System. By 1909he articulated his full vision of “One System, One Policy, Universal Service” See Annual Reportof the AT&T Co., 1907, p. 28; 1909, pp. 18-19. On the differences between the original andmodern notions of “universal service,” see Milton J. Mueller, Universal Service: Competition,Interconnection, and Monopoly in the Making of the American Telephone System (Cambridge,MA: MIT Press, 1997), ch. 1. H.C. Osborne predicted that Vail’s vision would finally be realizedca. 1935 with the implementation of Bell’s General Toll Switching Plan; H.C. Osborne, “AGeneral Switching Plan for Telephone Toll Service,” Bell System Technical Journal, 9 (July1930), 429-47; “Technical Developments Underlying the Toll Services of the Bell System,” BellSystem Technical Journal, 15 (July 1936, Supplement), 58-65; and Bancroft Gherardi and F.B.Jewett, “Telephone Communication System of the United States,” Bell System Technical Journal,9 (January 1930), 61-62. The General Switching Plan forged a truly national telephone networkthrough the establishment of additional regional centers – Atlanta, Dallas, Denver, San Francisco,and Los Angeles – to mediate intra- and interregional traffic.

3. Annual Report of the AT&T Co., 1908, 22.

4. See, for example, Joel A. Tarr, et al, “The City and the Telegraph: Urban Telecommunicationsin the Pre-Telephone Era,” Journal of Urban History, 14 (November 1987), 54; and morerecently, Bill Gates, Business @ the Speed of Light:

5. On the maturing US urban system, see Michael P. Conzen, “The Maturing Urban System inthe United States, 1840-1910,”Annals of the Association of American Geographers, 67(March1977), 88-108.

6. In similar fashion James A. Wheeler and Ronald L. Mitchelson use Federal Express shipmentsto delineate the hierarchy of metropolitan centers in the early 1980s; “Information Flows amongMajor Metropolitan Areas in the United States,” Annals of the American Association ofGeographers, 79 (December 1989), 523-43.

7. AT&T Archives, box 1259, AT&T Co., E.J. Hall, Long Distance Telephone Work, 1887. Tojustify their proposed acquisition of Western Union, Bell officials elaborated the complementaritybetween telegraph and telephone service; AT&T, Annual Reports (1909), pp. 31-32; (1910), 49-53; and (1911), pp. 41-43.

END NOTES

35

8. Alan J. Moyer, “Urban Growth and the Development of the Telephone: Some Relationships atthe Turn of the Century,” in Ithiel de Sola Pool, ed., The Social Impact of the Telephone(Cambridge, MA: MIT Press, 1977), 357-61; Claude S. Fischer, America Calling: A SocialHistory of the Telephone to 1940 (Berkeley: University of California Press, 1992), esp. ch. 7,document the narrow scope of telephone communications before 1940. Even in the 1970s,telephone use by residential customers exhibits the same spatial patterns; see J.W. Simmons,“Interaction among the Cities of Ontario and Quebec,” in L.S. Bourne and R.D. McKinnon, eds.,Urban Systems Development in Central Canada: Selected Papers (Toronto: University of TorontoPress, 1972), 203-06; and Fischer, America Calling, 225-26.

9. In the mid-1920s the time to complete local connections varied from 19 seconds in smallercities to 29 seconds in larger ones; AT&T Company, Proceedings of the Bell EducationalConference June 21-25, 1926 (New York: AT&T Co., 1926), 55; Gherardi and Jewett,“Telephone Communication System,” 7; U.S. Bureau of the Census, Special Report, Telephones:1907 (Washington, D.C.: G.P.O., 1910), 84. Bell estimates indicate that the longest componentwas the time spent waiting for the desired party to answer the call.

10. This scenario fits all urban exchanges, except those located in the largest cities, which wereserved by multiple central offices and switchboards connected by trunk lines. In the latter settingstechnology and operating methods more closely resembled those of toll service. It also presumesthe adoption of the common battery system, whose central power source automatically signaledoperators of the demand for and termination of a connection. For an excellent description ofurban exchange technology and switching methods, see Milton L. Mueller, “The SwitchboardProblem: Scale, Signaling, and Organization in Manual Telephone Switching, 1877-1897,”Technology and Culture, 30 (July 1989), 534-60; and Kenneth Lipartito, “When Women WereSwitches: Technology, Work, and Gender in the Telephone Industry, 1890-1920,” AmericanHistorical Review, 99 (October 1994), 1075-1111. M.D. Fagen et al, A History of Engineeringand Science in the Bell System, The Early Years (1875-1925) (Warren, NJ: Bell TelephoneLaboratories, 1975), provides a comprehensive overview of Bell technology and operatingmethods during this period.

11. By this time, Bell telephone operators were almost exclusively women; see Lipartito, “WhenWomen Were Switches.”

12. AT&T Company, Annual Report, 1907, 22; Bell System Educational Conference, 59;Gherardi and Jewett, “Telephone Communication System,” 34; and Fagen et al, History ofEngineering and Science, 619-22. In 1926 the average speed of toll service varied from 35seconds for calls to nearby central offices with direct trunk connections to over 12 minutes forlong-haul traffic over 250 miles. For “regular” toll board service, the average was 7 minutes.

13. The design and operation of the toll plant derived from two sources: the greater technologicalsophistication and cost of switching and transmission equipment and the limited scope of demand– the infrequent use of toll service by most customers and shorter duration of calls; F.P.Valentine, “Problems in Telephone Traffic Engineering,” Telephony, 61 (July 15, 1911), 76.

36

These points are discussed more fully in sections two and three below.

14. The source of these data are U.S. Department of Commerce, Bureau of the Census,Historical Statistics of the United States: Colonial Times to 1970, Part 2, Chapter RCommunications, Series R9-12.

15. Traffic exceeded average monthly levels during the fall harvest season, and was belowaverage during the summer months; Frank F. Fowle, “Economical Development of TollTerritory,” Telephony, 17 (December 19, 1908), 638. Recognizing the connection betweenaggregate economic activity and toll service, Bell’s Commercial Engineering Department devotedconsiderable resources to analyzing and forecasting aggregate trends; W.C. Helmle, “The Relationbetween Telephone Development and Growth and General Economic Conditions,” Bell SystemQuarterly, 4 (January 1925), 8-21; and F.P. Valentine, “Some Phases of the Commercial Job,”Bell System Quarterly, 5 (January 1926), 34-43.

16. F.E. Richter, “The Telephone as an Economic Instrument,” Bell System Quarterly, 4(October 1925), 281-95.

17. S.D. Levings, “The Development Study,” Telephony, 17 (March 13, 1909), 306, observedthat small retailers often purchased limited service or dispensed with telephone service entirely. Similarly, in New Orleans, retailers grocers and druggists favored dual or competitive service,even if it fragmented the urban market; “The Telephone Situation in New Orleans,” AmericanDaily Telephone, 17 (May 12, 1908), 406.

18. AT&T Co., Bell System Educational Conference, 59. Fowle, “Economical Development,”636, made a similar point: “Leased wires are used extensively by stock and grain brokers, packingcompanies, press associations and large industrial concerns.”

19. AT&T Archives, Box 1057, AT&T Co., T.B. Doolittle, Toll Lines, 1904. The analysis inthis section draws mainly on Doolittle’s manual. Subsequent quotations and references, unlessnoted otherwise, are to this document. Prior to Vail’s return to the helm of AT&T, the efforts toachieve this goal were frustrated by the myopic vision of managers of the parent company and itslicensed operating companies. For example, Doolittle’s “procedure” was first adopted by a Belloperating company – the Missouri and Kansas – in 1905, and as late as July 1906 he was writingto the head of the engineering department on ways to interest the other operating companies inadopting Bell standards; AT&T Archives, box 2020, Thomas B. Doolittle, Annual Reports, 1903,1906-1908, Doolittle-Fish, 14 January 1907. See Robert W. Garnet, The Telephone Enterprise:The Evolution of the Bell’s System Horizontal Structure, 1876-1909 (Baltimore: The JohnsHopkins University Press, 1985), 113-25 and 131-46; Kenneth Lipartito, The Bell System andRegional Business: The Telephone in the South, 1877-1920 (Baltimore: The Johns HopkinsUniversity Press, 1989), 158-67; Louis Galambos, “Theodore N. Vail and the Role of Innovationin the Modern Bell System,” Business History Review, 60 (Spring 1992), 95-126.

37

20. AT&T Co., Annual Report, 1907, 22-23. From this perspective, distance is an economicconcept and is more accurately gauged by the facility of communications, such as the averagenumber of paths connecting two places or the speed of service between them; see Ronald Abler etal, Spatial Organization: The Geographer’s View of the World (Englewood Cliffs: Prentice-Hall,1971), 257-63; Simmons, “Interaction among Cities,” 212-15; Puffert, “The Economics of SpatialNetwork Externalities,” 33; and Michael D. Irwin and John D. Kasarda, “Air Passenger Linkagesand Employment Growth in U.S. Metropolitan Areas,” American Sociological Review, 56(August 1991), 528.

21. The commercial application of loading and repeating (or amplification) technology in the1920s obviated this problem; Gherardi and Jewett, “Telephone Communication System,” 33-34;Osborne, “Technical Developments,” 13-22; and Fagen et al, History of Engineering and Science,251-56.

22. Although the map depicts a small, rather simple network, it does not differ in kind from morecomplex systems around larger metropolitan centers; see “United States Telephone Circuit Mapof Ohio,” Telephony, 16 (November 28, 1908), 535; “Lincoln Telephone & Telegraph to MakeImportant Extensions This Year,” Telephony, 17 (April 17, 1909), 472; “Circuit Map ofIndependent Lines in Michigan,” Telephony, 60 (June 17, 1911), 712; and AT&T Archives, Box117-04-01, AT&T Co., Telephone Line Atlas.

23. This comment amounts to altering the index by weighting the number of switches for eachroute by its share of the center’s total traffic. As inspection of panel B in Table 4 clearlyindicates, this modification would reinforce the nodality of center A and D.

24. Selecting auxiliary routes involved a trade-off between the additional time sending calls alongmore roundabout paths and the waiting time for a clear circuit on the primary route. A secondpath offered a real alternative when primary circuits were busy; a tertiary path was used only incase of an emergency, such as damage to primary lines. All exchanges had practically the samerange of second routes, but the availability of a third route critically depended on a center’snodality.

25. AT&T Archives, Box 1057, T.B. Doolittle, Toll Lines, 1904. See also Box 1359, AT&TCo., Engineering Department, Bulletins, 1902, The Center Checking System; and Box 2020,AT&T Co., Thomas B. Doolittle, Annual Reports, 1903, 1906-1908, Doolittle-Fish, 14 January1907. In his 1906 annual report Doolittle specifies the maximum distance of a tributaryconnection to be 30 miles.

26. F.P. Valentine, “Problems in Telephone Traffic Engineering,” Telephony, 61 (July 1, July 8,July 15, July 22), 26-27, 47-48, 75-76, 105-06.

27. AT&T Archives, Box 1057, AT&T Co., T.B. Doolittle, Toll Lines, 1904; and Box 2020,Thomas B. Doolittle, Annual Reports, 1903, 1906-1908, Doolittle-Fish, 7 March 1904.

38

28. AT&T Archives, Box 1057, AT&T Co., T.B. Doolittle, Toll Lines, 1904; and Box 1359,Engineering Department, Bulletins, 1902, The “Center Checking” System.

29. AT&T Co., T.B. Doolittle, Toll Lines, 1904.

30. In this case the combined traffic between center A and centers D and F justify the directcircuit to D. By viewing D and F as a single large center, the volume of direct traffic accounts for80 percent of the total.

31. Jacob Price, “Economic Function and the Growth of American Port Towns,” Perspectives inAmerican History, 8 (1974), 129-30. For a similar notion of entrepots as an “intelligencecomplex” or control and coordination center, see James E. Vance, Jr., The Merchant’s World:The Geography of Wholesaling (Englewood Cliffs: Prentice-Hall, 1970), 149-50; and David R.Meyer, “A Dynamic Model of the Integration of Frontier Urban Places into the United StatesSystem of Cities,” Economic Geography, 56 (January 1980), 121-22.

32. AT&T Archives, box 1011, AT&T Co., Building of Early Long Distance Lines, Hall-Hudson, 21 January 1888.

33. Jeffrey Rohlfs, “A Theory of Interdependent Demand for a Communications Service,” BellJournal of Economics and Management Science, 5 (Spring 1975), 30-31, uses the term“community of interest groups’ to characterize interdependent demands for telephone service.

34. I have borrowed the term “eco-technic” from Ross Thomson, “The Eco-Technic Process andthe Development of the Sewing Machine,” in Gavin Wright and Gary Saxanhouse, eds.,Technique, Spirit, and Form in the Making of Modern Economies: Essays in Honor of William N.Parker (Greenwich, CT: JAI Press,), 243-69, who emphasizes the dialectic between technologyand the market in the process of innovation. For similar notions, see Thomas P. Hughes, “TheEvolution of Large Technological Systems,” in Wiebe E. Bijker, Thomas P. Hughes, and Tevor J.Pinch, eds., The Social Construction of Technological Systems: New Directions in the Sociologyand History of Technology (Cambridge, MA: MIT Press: 1987), 51-82; and William Lazonick,Business Organization and the Myth of the Market (New York: Cambridge University Press,1991). I am grateful to Kenneth Lipartito, who cautioned against reifying the economicrelationships embodied in complex technological systems, like the telephone network, into thevery instruments or “artifacts” themselves.

35. They could also reduce circuit quality by using lower gauge copper wire or even bysubstituting iron for copper; see, for example, Southern Bell Telephone and Telegraph Co.(SBT&T), Annual Report, 1907 (Atlanta: n.p., n.d.), Pickernell-Hall, 21 February 1908. Circuitsmade of lower gauge copper and iron were effective only on low density, short haul routes;Osborne, “General Switching Plan,” 443-44; Fagen et al, History of Engineering and Science,343.

36. AT&T Archives, Box 1309, Relation between Population and Rates, 1906, Ford-Fish, 24May 1906; Frank F. Fowle, “Economical Development of Toll Territory,” Telephony, 17 (January

39

9, 1909), 41. For example, increasing the number of circuits from 4 to 5 reduced the “averageannual charges” per circuit mile from $36.2 to $30.2 or by 16.6%.

37. These figures refer to feasible loads during peak periods, i.e., the maximum number of callssubject to tolerable levels of congestion or uncompleted connections because of busy circuits;AT&T Archives, box 1057, AT&T Co., T.B. Doolittle, Toll Lines, 1904; and R.I. Wilkinson,“Beginnings of Switching Theory in the United States,” Electrical Engineering, 75 (September1956), 796-98. See also Frank F. Fowle, “Economical Development of Toll Territory,”Telephony, 17 (January 2), 9; and George K. Gann, “Handling Long Distance Traffic,”Telephony, 18 (August 28, 1909), 206. According to Wilkinson, Bell engineers drafted “the first‘comprehensive’ traffic engineering” study in 1903, which demonstrated these “objectiveefficiencies” empirically. For contemporary estimates, see Leonard Waverman, “The Regulationof Intercity Telecommunications,” in Almarin Phillips, ed., Promoting Competition in RegulatedMarkets (Washington, D.C.: The Brookings Institution, 1975), 213-14.

38. Edward C. Molina, “Application of the Theory of Probability to Telephone TrunkingProblems,” Bell System Technical Journal, 6 (July 1927), 461-94; see also Wilkinson, “TheBeginnings of Switching Theory,” 796-803. The “normal” conditions, necessary to realize theseeconomies, are discussed more fully in the following note and below. The same principles applyto wholesale trade and explain the rationale for bulk transactions – the accumulation of orders andshipments over time and space.

39. A simple example, taken from William Feller, An Introduction to Probability Theory and ItsApplications, vol. 1 (New York: John Wiley, 1968), 191-92, illustrates the point. A city isdivided into two equal districts, each served by separate toll centers furnished with same numberof circuits, k. The likelihood that a caller will encounter a busy signal depends on whetherdemand at either center exceeds circuit capacity, or Prob(X1>k or X2>k), where X1, X2 = thedemand for toll calls at centers 1, 2 respectively. If the demand for calls in the two districts areessentially uncorrelated, then it is far less likely that both centers would encounter capacityconstraints at the same time, or Prob(X1>k and X2>k) << Prob(X1>k or X2>k). Therefore, whenone center experienced excess demand, the other, most likely, would have excess capacity –unutilized circuits that could handle the overflow traffic. By consolidating the two centers intoone or enlarging the circuit group to 2k, the company could either improve the grade of service(i.e., reduce the risk of congestion delays) or increase average circuit loads.

40. In his annual message to stockholders in 1912, Vail trumpeted this innovation: “At the end of[the year], there was a total of 54,750 miles of the heaviest gauge wires equipped with the newarrangement;” AT&T Co., Annual Report, 1912, 22. It not only doubled “their transmissionefficiency [range],” but yielded “circuits equivalent to 12,600 miles of the heaviest-gaugecircuits.” On the invention and development of loading, see Neil H. Wasserman, From Inventionto Innovation: Long-Distance Telephone Transmission at the Turn of the Century (Baltimore:Johns Hopkins University Press, 1985); and Fagen et al, History of Engineering and Science, 236-40.

40

41. According to Wasserman, From Invention to Innovation, 102-25, loading coils reduced thefixed costs of long distance transmission facilities by over 50 percent. Still, because of thecomplexity of the technology, its early adoption was restricted to high density cables. Loadingtechnology also spurred the introduction of carrier circuits, which transmitted multiple calls ondifferent frequency bands over a single physical circuit; see Osborne, “Technical Developments,”9-33; Gherardi and Jewett, “Telephone Communication System,” 45-57; and Fagen et al, Historyof Engineering and Science, 240-41.

42. The reference to “throughput” is intended to relate the organizational and operationalinnovations developed by Bell to similar processes adopted by other large-scale enterprises inproduction and distribution during the period; see Alfred D. Chandler, Jr., Scale and Scope: TheDynamics of Industrial Capitalism (Cambridge, MA: Harvard University Press, 1990), 21-31.

43. AT&T Archives, Box 1285, Toll Line Service, 1896-1898, Doolittle-Hudson, 25 May 1898. I am grateful to David Gabel for bringing this document to my attention.

44. When operated by a common power source, the method yielded even greater savings. Withthis technology, “supervising signals or lamps” on switchboards would indicate when circuitswere busy and in turn the completion of a connection. Toll operators could account moreprecisely for toll calls and rapidly “clear the trunk instantly for the next call, thus making possiblethe handling of a larger number of calls per trunk than at present.” The application of thecommon battery system to the toll plant depended on improvements in line quality and in trafficdensity and so was also limited to larger markets. See also Osborne, “Technical Developments,”48-49; Fagen et al, History of Engineering and Science, 488-502; and Mueller, “The SwitchboardProblem.”

45. AT&T Archives, Box 1285, Toll Line Service, 1896-1898, Doolittle-Hudson, 24 May 1898.

46. The editors of Telephony were particularly interested in improving toll operating methods,and the journal ran a series of articles describing these innovations and their economic benefits;see “Development of Long Distance Business,” Telephony, 13 (May 1907), 305-06; “ImprovingToll Line Service,” Telephony, 13 (May 1907), 311-12; “Efficiency in Toll Operating Methods,”16 (October 31, 1908), 433-34; Fowle, “Economical Development of Toll Territory,” Telephony,17 (December 12, 1908), 612-14, and 18 (January 2, 1909), 8-9; Gann, “Handling Long DistanceTraffic,” 205-07; “Methods of Securing High Efficiency in the Operation of a Toll Line System,”Telephony, 60 (May 6, 1911), 557-58; and Valentine, “Problems in Telephone TrafficEngineering.” See also Gherardi and Jewett, “Telephone Communication System,” 35-36;Osborne, “Technical Developments,” 35-40; Fagen et al, History of Engineering and Science,618-24.

47. See, for example, “What the Two-Number System Is--A Toll Operating Method forConcentrated Business Traffic,” Telephony, 61 (September 9, 1911), 308.

41

48. AT&T Archives, Box 1285, AT&T Co., Toll Line Service, 1892-1896, Doolittle-Davis, 4June 1896.

49. AT&T Archives, Box 1285, AT&T Co., Toll Line Service, 1892-1896, Doolittle-Davis, 4June 1896 (diagram); Box 1330, AT&T Co., Toll Line Service, 1900, Doolittle-Hudson, 20 June1900; AT&T Co., Toll Line Service, 1903-1904, Doolittle-Hudson, 22 March 1904; and Box2020, AT&T Co., Thomas B. Doolittle, Annual Reports, 1903, 1906-1908, Doolittle-Fish, 7March 1904.

50. AT&T Archives, Box 1285, AT&T Co., Toll Line Service, 1892-1896, Doolittle-Davis, 4June 1896; Doolittle-Davis, 7 July 1896; Box 1309, AT&T Co., Relation Between Population andRates, 1906, Ford-Fish, 24 May 1906; and Box 1330, AT&T Co., Toll Line Service, 1903-1904,Doolittle-Hudson, 22 March 1904.

51. This point is developed more fully in a companion paper; David F. Weiman, “Systemic Limitsto a National Telephone Network: The Anomaly of the Deep South” (unpublished manuscript,Queens College, 1996).

52. AT&T Archives, Box 1285, AT&T Co., Toll Line Service, 1897-1898, Doolittle-Hudson, 14June 1898.

53. Richard Whitcomb, “The Key Town Plan of Selling by Telephone,” Bell System Quarterly, 8(January 1929), 52. Other examples include the use of the telephone in facilitating the purchaseand sale of standard commodities, like coal and other materials, and in dispatching trains; see “MyPartner, the Telephone – Using Modern Methods,” Telephony, 60 (April 22, 1911), 490-22; G.K.Heyer, “Telephone Economy in Railway Service,” Telephony, 61 (January 1911), 19-21; andRichter, “Telephone as an Economic Instrument,” 286-90.

54. Whitcomb, “Key Town Plan,” 52; Richter, “Telephone as an Economic Instrument,” 293.

55. AT&T Archives, Box 1342, AT&T Co., Rural Telephone Service, 1903-1904, Allen-Fish, 12August 1903. Toll service also transformed how farmers marketed their crops. According to oneobserver, “the rural telephones and rural mail routes ... put the street grain broker out of businesscompletely,” as farmers negotiated the sale of their crops over the telephone before they cartedthem to town; “The Rural Telephone and the Grain Market,” Telephony, 16 (September 19,1908), 263-64.

56. Price, “Economic Function,” 129, 142.

57. Fowle, “Economical Development,” 9; and AT&T Archives, Box 1309, AT&T Co., RelationBetween Population and Rates, 1906, Wray-Hibbert, 14 May 1906, Ford-Fish, 24 May 1906. Specifically, where demands are correlated, “but not fully coincident ... it is possible to exploit ...economies of scope;” Bridger M. Mitchell and Ingo Vogelsang, Telecommunications Pricing:Theory and Practice (New York: Cambridge University Press, 1991), 11.

42

58. Molina, “Telephone Trunking Problems,” 467-68; and Wilkinson, “Beginnings of SwitchingTheory,” 800-01. Referring to the simple example in note 37 above, these conditions insure thatthe probability of congestion is much smaller in the case of consolidation; i.e., Prob(X1>k orX2>k) = Prob(X1>k) + Prob(X2>k) - Prob(X1>k)@Prob(X2>k) . Prob(X1>k) + Prob(X2>k) =2Prob(X1>k) > Prob(X1+X2>2k).

59. Roger I. Wilkinson, “Theories of Toll Traffic Engineering in the U.S.A.,” Bell SystemTechnical Journal, 35 (March 1956), 443-46. See also A.E. Joel, Jr. et al, A History ofEngineering and Science in the Bell System: Switching Technology (1925-75) (Warren, NJ: BellLaboratories, 1982), 95-100; and Hills, Telecommunications Switching Principles, ch. 4.

60. Fowle, “Economical Development,” 638.

61. For example, local processing industries not only enhanced the value of farm produce, butoperated during the non-peak seasons in agriculture, when the network’s capacity was otherwiseunderutilized. Additionally, improved communications networks fostered the spatial dispersion ofprocessing facilities nearer to sources of supply, and so contributed to the formation of theseregional complexes. Under these conditions, Berry’s entropy model of urbanization implies auniform rank-size distribution; see Brian J.L. Berry, “City Size Distribution and EconomicDevelopment,” in John Friedmann and William Alonso, eds., Regional Development and Planning:A Reader (Cambridge, MA: MIT Press, 1964), 147-49; and Harry W. Richardson, “Theory of theDistribution of City Sizes: Review and Prospects,” Regional Studies, 7 (September 1973), 244. My formulation also satisfies the criticism of Berry’s model by Robert Higgs, “Central PlaceTheory and Regional Urban Hierarchies: An Empirical Note,” Journal of Regional Science, 10(??? 1970), 253.

62. AT&T Archives, Box 1309, Relation between Population and Rates, Ford-Fish, 24 May1906 (emphasis added).

63. In fact, geographers use this very evidence to delineate urban systems and the trade areasof higher order centers; John R. Borchert and Russell B. Adams, “Trade Centers and Trade Areasof the Upper Midwest,” Upper Midwest Economic Study, Urban Report No. 3 (September1963); and Ronald Abler, “The Telephone and the Evolution of the American Urban System,” inIthiel de Sola Pool, ed., The Social Impact of the Telephone (Cambridge, MA: MIT Press, 1977),318-41. According to Abler et al, Spatial Organization, 265-66, a more nodal or dominant centersends its greatest flow of traffic to a smaller city, where size is measured by population.

64. David F. and Richard C. Levin Weiman, “Preying for Monopoly? The Case of Southern BellTelephone Comapny, 1894-1912,” Journal of Political Economy, 102, (February1994), 103-26;see also John V. Langdale, “The Growth of Long-Distance Telephony in the Bell System,1875-1907,” Journal of Historical Geography, 4 (April 1978), 145-59; Kenneth Lipartito, TheBell System and Regional Business: The Telephone in the South, 1877-1920 (Baltimore: JohnsHopkins University Press, 1987), 82-100; Milton L. Mueller, “The Telephone War:Interconnection, Competition, and Monopoly in the Making of Universal Telephone Service,

43

1894-1920” (Ph.D. dissertation, University of Pennsylvania, 1989), 177-81; and David Gabel,“Competition in a Network Industry: The Telephone Industry, 1894-1910,”Journal of EconomicHistory, 54, (Sept. 1994), 543-72.

65. The threat of competition from merchants in nearby towns frequently prompted merchants todemand greater toll connections; see AT&T Archives, Box 1214, AT&T Co., Rural TelephoneService, 1899-1902, President of Nebraska Telephone-Fish, 14 April 1902.

66. To be fair, Doolittle was making a different point; toll traffic depended more on the totalpopulation of a center than the number of exchange subscribers. Managers, he worried,myopically focused on the latter, and so underestimated the potential toll traffic in their territory,especially where competition was strong. Still, his empirical evidence shows that toll receipts percapita varied from a minimum of $0.262 in a city of 4703 people to $1.610 in a town of 1342;AT&T Archives, Box 1057, T.B. Doolittle, Toll Lines, 1904. Also, in an earlier document, heoffered an important qualification: “it is fair to presume that there should be some relation in theamount of receipts per inhabitant, provided that the toll facilities are of like character, andadequate;” AT&T Archives, Box 1285, AT&T Co., Toll Line Service, 1897-1898, Doolittle-Hudson, 3 June 1898.

67. AT&T Archives, Box 1309, AT&T Co., Relation between Population and Rates, 1906,Wray-Hubbard, 14 May 1906 (quotation); Ford-Fish, 24 May 1906.

68. AT&T Archives, Box 1285, AT&T Co., Toll Line Service, 1892-1896, Doolittle-Davis, 8July 1896; AT&T Co., Toll Line Service, 1897-1898, Doolittle-Hudson, 14 June 1898; Box 2020,AT&T Co., Thomas B. Doolittle, Annual Reports, 1903, 1906-1908, Doolittle-Fish, 4 January1907; Box 2026, SBT&T Co., Toll Traffic Matters, 1909, Doolittle-Carty, 14 July 1909. Geographers define the hierarchical order of a center by the number and geographic range of itseconomic functions; it is often highly correlated with population. See, for example, Brian J.L.Berry, Geography of Market Centers and Retail Distribution (Englewood Cliffs: Prentice-Hall,1967), 14-18, 35-40.

69. AT&T Archives, Box 2020, AT&T Co., Thomas B. Doolittle, Annual Reports, 1903, 1906-1908, Doolittle-Fish, 7 March 1904.

70. Whitcomb, “Key Town Plan,” 52-53.

71. Richter, “Telephone as an Economic Instrument,” 291.

44

TABLE 1TELEPHONE TRAFFIC BY BUSINESS CLASS: BUFFALO, NY, 1891

“BUSINESS CLASS” Local Toll and Long Distance

Stations Calls Calls/Station

Stations % Toll Calls Calls/Station

Public Stations 5 351 70.2 5 100.0 246 49.20

Hotels 27 5151 190.8 13 48.1 292 22.46

Telegraph 8 3716 464.5 5 62.5 71 14.20

Commission Merchants 55 6757 122.9 34 61.8 225 6.62

Grain Dealers 58 10763 185.6 30 51.7 162 5.40

Street Railroads 7 921 131.6 3 42.9 16 5.33

Lumber 63 8435 133.9 36 57.1 190 5.28

Coopers and Materials 4 545 136.3 4 100.0 21 5.25

Banking 27 6045 223.9 17 63.0 81 4.76

Builders 44 3940 89.5 24 54.5 112 4.67

Shipping 30 11483 382.8 28 93.3 128 4.57

Metals 59 8389 142.2 33 55.9 148 4.48

Hospitals & Charitable Orgs. 20 2073 103.7 9 45.0 33 3.67

Grocers 67 6005 89.6 35 52.2 126 3.60

Miscellaneous 57 4756 83.4 26 45.6 92 3.54

Real Estate 49 3525 71.9 16 32.7 54 3.38

Hardware 58 6170 106.4 36 62.1 118 3.28

Coal 66 11158 169.1 44 66.7 144 3.27

Caterers 23 1426 62.0 4 17.4 13 3.25

Druggists 83 11439 137.8 53 63.9 169 3.19

Butchers/Fish Stores 49 4748 96.9 23 46.9 71 3.09

Newspapers 12 2851 237.6 11 91.7 33 3.00

Oil (Petroleum) 11 2272 206.5 5 45.5 15 3.00

Paper 13 911 70.1 8 61.5 19 2.38

Electrical Business 15 1581 105.4 6 40.0 14 2.33

Liquors 51 4665 91.5 22 43.1 51 2.32

Residences 289 14110 48.8 53 18.3 119 2.25

Paving & Materials 19 3229 169.9 10 52.6 22 2.20

Customs & Revenue 5 785 157.0 1 20.0 2 2.00

Undertakers 19 1949 102.6 5 26.3 10 2.00

TABLE 1TELEPHONE TRAFFIC BY BUSINESS CLASS: BUFFALO, NY, 1891

“BUSINESS CLASS” Local Toll and Long Distance

Stations Calls Calls/Station

Stations % Toll Calls Calls/Station

45

Railroads 71 29560 416.3 30 42.3 56 1.87

Doctors 144 5966 41.4 20 13.9 34 1.70

Dry Goods 23 1979 86.0 9 39.1 15 1.67

Stables 59 4182 70.9 31 52.5 51 1.65

Law 63 6189 98.2 37 58.7 57 1.54

Leather & Rubber 21 1432 68.2 15 71.4 22 1.47

Paints & Oils 16 1060 66.3 3 18.8 4 1.33

Ice 7 588 84.0 3 42.9 4 1.33

City Offices 21 4329 206.1 22 104.8 28 1.27

Carriage Builders 15 876 58.4 5 33.3 6 1.20

Stationers 23 1749 76.0 7 30.4 6 0.86

Car Builders 4 1407 351.8 3 75.0 2 0.67

Express & fast freight 13 1533 117.9 4 30.8 2 0.50

Gas 6 867 144.5 6 100.0 3 0.50

Furniture 24 1922 80.1 7 29.2 2 0.29

Insurance 26 2683 103.2 14 53.8 4 0.29

Teaming and Carting 5 431 86.2 0 0.0 0

Bell Tel. (Office) 16 7015 438.4 0 0.0 0

Business 1412 203490 144.1 737 52.2 2694 3.66

Wholesale 405 93068 229.8 237 58.5 1349 5.69

Residential 433 20076 46.4 73 16.9 153 2.10

Professionals 144 5966 41.4 20 13.9 34 1.70

Other 289 14110 48.8 53 18.3 119 2.25

Public Stations 5 351 70.2 5 100.0 246 49.20

Total 1850 223917 121.0 815 44.1 3093 3.80

Notes: Stations simply mean telephone units. % Toll equals the percentage of exchange subscribers making tollcalls. Public stations are telephones located in public places, such as stores.

Source: AT&T Archives Book Collection, Telephone Switchboard Conference, March 15-18, 1892.

46

TABLE 2TOLL CALLS IN AN UPSTATE NEW YORK EXCHANGE, JULY 1900

A. Toll Calls by Type of Customer

Subscribers Toll UsersToll Calls Toll Calls Per

All Subscribers Subscriber Toll Users

Total 490 250 1608 1320

Business 275 176 1165 1165 4.24 6.62

% of Total 56.1 70.4 72.5 88.3

Residential 215 74 155 155 0.72 2.09

% of Total 43.9 29.6 9.6 11.7

Public Stations 2 288

% of Total 17.9

B. Size Distribution of Toll Calls

Number of CallsBusiness Residential

% Subscribers % Calls % Subscribers % Calls

0 36.0 0.0 65.6 0.0

1-5 42.9 21.5 32.6 73.5

6-10 11.3 20.6 0.9 11.6

11-20 5.1 17.5 0.9 14.8

20+ 4.7 40.3

Source: AT&T Archives, Box 1330, AT&T Co., Toll Line Service, 1901, Doolittle-Cochrane1/16/1901

47

TABLE 3LARGEST USERS OF LONG DISTANCE SERVICE, NEW YORK CITY, 1926

Name “Business Class”Messages perBusiness Day

Pennsylvania Hotel Hotel 197

Commodore Hotel Hotel 156

Biltmore Hotel Hotel 93

McAlpin Hotel Hotel 89

Walldorf-Astoria Hotel Hotel 87

Astor Hotel Hotel 69

Roosevelt Hotel Hotel 62

Vanderbilt Hotel Hotel 49

Belmont Hotel Hotel 45

Chelsea Bro. and Robbins Inc. Fish 35

Coleman and Reitze Bankers 30

Shelton Hotel Hotel 29

Prince George Hotel Hotel 27

McAlpin Hotel Hotel 26

Imperial Hotel Hotel 23

Ford Motor Co. of Delaware Automobile 23

Standard Oil Co. of New York Oil 22

United States Rubber Rubber 22

First National Corp. of Boston Bankers 18

Times Square Hotel Hotel 18

General Electric Electric 18

Woodstock Hotel Hotel 18

Ambassador Hotel Hotel 17

Barrett Company Roofing Materials 17

Alamac Hotel Hotel 17

Breslin Hotel Hotel 16

Bethlehem Steel Co. Steel 16

Yale Club of New York Club 15

Debevoise Anderson Co., Inc. Iron 15

W.J. Rainey, Inc. Coal 15

Langham Hotel Hotel 15

SOURCE: AT&T Co., Proceedings of the Bell System Educational Conference, June 21-25, 1926 (NewYork: AT&T Co., 1926), 60.

48

TABLE 4TRAFFIC PATTERNS IN AN REGIONAL TOLL NETWORK

A. Patterns of Calls by Center

CenterNo. ofCalls

AverageSwitches

Alternative Routes Concentrationof Calls2nd 3rd

A 7947 0.83 87.0 56.5 57.5

D 5536 0.70 87.0 39.1 60.4

I 1920 1.43 87.0 26.1 73.5

J 3306 1.04 100.0 78.3 72.9

K 885 2.00 100.0 4.3 79.4

W 2150 1.78 87.0 56.5 97.3

B. Patterns of Calls by Type of RouteCenter Type of

Route% Average

Calls% of Calls

A Direct 34.8 850.8 85.6

One 47.8 83.5 11.6

Two+ 17.4 55.5 2.8

D Direct 39.1 555.6 90.3

One 52.2 42.5 9.2

Two+ 8.7 13.0 0.5

J Direct 21.7 524.6 79.3

One 52.2 53.8 19.5

Two+ 26.1 6.3 1.1

I Direct 17.4 335.5 69.9

One 26.1 85.2 26.6

Two+ 56.5 5.2 3.5

K Direct 13.0 220.3 74.7

One 17.4 21.8 9.8

Two+ 69.6 8.6 15.5

W Direct 4.3 1235.0 57.4

One 30.4 108.4 35.3

Two+ 65.2 10.4 7.3

Notes: Number of calls equals the volume of monthly traffic. Average switches equals the average numberof relays required to reach any destination in the network. Alternative routes measure the percentage ofdestinations, furnished with a secondary (2nd) and tertiary (3rd) route. Concentration of calls equals theshare of traffic on the largest four routes. Type of route indicates the number of switches: direct or none,one intermediate switch, and at least two switches.

SOURCE: AT&T Archives, Box 1057, T.B. Doolittle, Toll Lines, 1904.

49

50