1.1. Meaning of functional

There are multiple ways the term functional has been used. These diverse ways confuse and blend different concepts when creating a typology. In general, function refers to what the town does for its residents, immediate surrounding area, region, state, or nation. Function is tied to why a settlement exists. The first principle commonly used is qualitative in nature and usually refers to cultural importance. A town is functionally important because it is a political, religious, or educational center. However, function does not automatically mean importance or purpose.

The term functional, in urban geography, urban sociology, and regional economics, rests on two interrelated constructs, both of which are tied to economics. First is the concept of central place. Walter Christaller first presented central place theory. This theory posits a size hierarchy of places related to the number of establishments (functions) present, which influences the spatial distribution of settlements. The smaller places (lower-order) have fewer functions than larger places (higher-order). There are three interrelated concepts: function, which is any establishment or institution that serves a population; a range, which is the maximum distance people are typically willing to travel to obtain a function; and threshold, which is the minimum size of an agglomeration of people necessary before a function is provided [2]. Each settlement has a tributary area for which it provides services. The larger the settlement, the larger the tributary or service area.

Edward Ullman explained it this way. “Services performed purely for a surrounding area are termed “central” functions by Christaller, and the settlements performing them “central” places. An industry using raw materials imported from outside the local region and shipping its products out of the local area would not constitute a central service.” [3].

Christaller’s definition of centrality (the excess of functions provided by a settlement above what the place’s population needs but is for the tributary region) plugs right into the construct of the economic base, the second principle underlying functional classifications.

Thus, any establishment, such as a grocery store, a hardware store, a library, or a manufacturer of automobiles, is a function. The question for central place theory is if it functions only for the settlement, the settlement’s hinterland, or outside its zone of influence. For the economic base, the question is what share of the function serves only the settlement’s population and what share serves the population outside the community. Christaller’s central place activity would be the same as a basic activity.

Non-basic economic activity refers to the activity which meets the needs of the place’s inhabitants, the local population. The basic component of economic activity of a settlement refers to that portion of the economy that brings in revenue from outside the area either by directly exporting goods or by persons coming into the community from outside its boundaries [4]. It is “basic” to the settlement’s existence. This partially overlaps the central place functions but extends them to areas external to the settlement and its tributary area.

Theoretically, it is necessary to determine the split between basic/non-basic for every establishment. For example, a grocery store will have customers from within and outside the town. This is very impractical. However, the basic/non-basic concept underlies all functional classifications, explicitly or implicitly. As John Alexander noted: “… any functional classification of settlements can be based on the basic functions which do differ from city to city without being confused with functions which are not much different.” [5]

These different uses of the term functional and the blending of the definitions can be found in the typologies created in the past.

1.2. Meaning of Urban

Both a theoretical and an operational issue, the question of what urban is has plagued social research due to its consequences on the units of analysis, thus affecting the structure of a study. There are two aspects of “urban” or the “city”: the actual physical footprint of the city and its labor market. The urban area’s physicality is the city’s core underlying perspective, and researchers try to approximate the physical city as best as possible. To be urban, usually, three criteria must be met: population size, density, and physical compactness or contiguousness.

Much research has been on the more significant incorporated places, such as the incorporated City of Chicago, the incorporated City of New York, or the incorporated City of Washington DC. In the past, the legally incorporated area of a city did match the physical city reasonably well. However, continued population growth has far outstripped the political boundaries of most cities.

The best approximation of settlement patterns on the ground is the Urban Area as defined by the U.S. Bureau of the Census. It is the best systematic attempt to identify a physical city empirically. Urban Areas ignore all political boundaries. An urban area of 50,000 in population or more (previously termed an Urbanized Area) is the core of all Standard Metropolitan Areas [6]. The Metropolitan Statistical Area (MSA) is a much better approximation of the economic “city” and its labor market, but it can include a large amount of undeveloped land or agricultural land. For example, the Tulsa, Oklahoma metro area has 6,460 square miles, but only 436 square miles are in urbanized areas. The MSA was not designed to account for the physical aspect of a city; it is more of a representation of the labor market and commuting area of the core urban area: its economic footprint.

The metropolitan area becomes a good approximation of the core settlement (the urban area), and its central place hinterland is the balance of the linked surrounding county areas.

1.3. Purposes

1.4. Approaches used in the past

Aurosseau made one of the earliest systematic efforts at an economic functional classification in 1921. He points out that there is usually one activity that overshadows the rest. He proposed a six-category typology: administrative, defense, culture, production, communication, and recreation [16].

As explained by Ergon Bergel, “… McKenzie bases his classification on economic stages: primary production (agriculture, husbandry, fishing, mining, and lumbering); the second, distribution of primary products; and the third, the manufacture of these products. The fourth category is residual, comprising all noneconomic activities.” [17] Bergel further elaborates these four primary types into five groupings: “… (a) primary “extraction,” mining, fishing, oil production; (b) conversion of raw materials into other goods, manufacture or industry; (c) spatial distribution of goods, transportation; (d) social distribution of goods, wholesale and retailing; or (e) supplementary services, financing, brokerage, advertising, and the independent professional services of lawyers, architects, accountants, and inventors.” [17]

Homer Hoyt presented a classification in 1941. He focused on what needs to be examined in an economic analysis by organizations locating facilities. It was a basic/non-basic employment analysis. The typology used was manufacturing, trade and finance, extraction, resorts, political capitals, higher education, and transportation. All other employment categories, such as retail and most services, were considered non-basic employment. [18]

Chauncy Harris developed a typology in 1943 based more on empirical data than previous systems, though he combined empirical data with the subjective [19]. His classification was based primarily on employment data from the U.S. Census but was supplemented with data on occupation. Other data, such as university enrollment, were also considered. When determining which category a city would be in, Harris only considered the employment as a percentage of the employment in Manufacturing, wholesaling, and retailing, not the city’s total employment. The percentages varied from industry to industry.

In a study of cities in the American South, John Hart replicated Harris’s classification, though he used only occupational data [20] to explore how cities changed over time. Ten occupational groups were used, and percentages were based on the total labor force. The cut-off points were set to the upper decile for each category. These values were set slightly higher for wholesale, educational, and mining centers and slightly lower for professional centers. For military centers, the cut-off was set to ten percent of the entire population. Cities not meeting these cut-offs were deemed diversified. Of interest, Hart considered the minimum percentages for each employment category as the minimum required by cities to be viable [20]. This minimum would be classed as non-basic employment.

Victor Jones, in 1953, presented a multifaceted classification of cities with over 10,000 in population [21]. While metro areas were mentioned, no data or classification was presented for them. He first determined the city’s status as the central city, a suburb, or independent. These were then subclassified as being employing, balanced, or dormitory. He also presented a classification based on rents and another based on the age of the housing stock. The classification of the economic base used multiple data sources: the Census of Manufactures, the Census of Business, and the Decennial Census. The study used labor force. The focus was on manufacturing and trade industries, with several employment categories excluded, particularly all the service industries, though services are included elsewhere. The selection criteria varied for each category without explaining the basis for the cut-off percentages used. The base for calculating the percentages changed from one category to another.

The classification developed by Howard Nelson in 1955 is the most well-known and the most widely used. Ten categories were left out: agriculture, forestry, fisheries, construction, utilities, business services, and private households. Military employment was included in public administration. Proportions were based on the total labor force and not the total employed. The criteria were applied uniformly across the employment categories. In choosing which industries were significant, Nelson used standard deviations. An industry had to be at least one standard deviation from the overall industrial mean. The classification of cities indicated which standard deviation applied: Mf, Mf2, or Mf3. Cities with no activity category one standard deviation or higher were classified as diversified. [22] The system allowed for multiple specializations.

Bergel, in 1955, presented a broad classification, utilized a combination of empirical and subjective criteria, and attempted to make the typology more universal and international [17]. Bergel combined economic, cultural, and political factors. The top-level categories were economic centers, political centers, cultural centers, recreational centers, residential cities, retirement cities, symbolic cities, and diversified cities.

In an exhaustive analysis by Otis Duncan and others, the attempt was to blend and merge economic base concepts, functional classification, and central place theory. The system used was extraordinarily complex and very data intensive. They relied heavily on input-out analysis tables. Underlying the system were two principles: a relationship to resource extraction and a market for outputs. They produced twelve categories within which Census employment categories were fitted. [23]

Categories two through seven were all manufacturing so a larger categorization would be Extraction, Manufacturing, and Services. The categories they used make a comparison to any other typology impossible. Additionally, they introduce a population size hierarchy into the analysis. Analysis was based on location quotients.

A second classification was introduced, which had little explanation on how the classification was constructed other than that value-added, wholesale receipts, and other Federal Reserve data were used. Metropolitan functions appear to be trade functions. [23]

Thomas Wilkerson created an interesting classification of Japanese cities in 1964 [12]. Several issues of international functional classification become apparent, such as differences in available data and definitions of what urban is. Wilkerson examined the male labor force in incorporated cities in 1920, 1930, 1950, and 1955. He developed six categories. The criteria for determining which Class a city was in were different from those used by other researchers: Both mining and agriculture had to have at least 25% of the male labor force employed; for industrial, the criteria was the percent employed in an industry divided by the sum of the percents of commerce, transportation, and administrative employment; for commerce, the criteria was the percent employed in commerce divided by the sum of the percents in industrial, transportation and administrative employment; for transportation and communication the criteria was the percent employed in transportation divided by the sum of the percents in industrial, commerce, and administrative employment; and for administrative and services, the criteria was the percent employed in transportation divided by the sum of the percent of industrial, commerce, and transportation employment. If an individual city had a ratio for a category one or more times found for the total urban employment, it was assigned as a specialty. [12] His system allowed for multiple specializations.

Fig 1 summarizes the classifications from the above studies.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Historical Functional Classifications.

https://doi.org/10.1371/journal.pone.0334284.g001

The discussion of past classifications highlights several problems/issues with functional classifications.

• Some employment categories were excluded.
• Labor force was used, not total employment.
• Employment data was based on occupational data.
• The units of analysis were limited to just large cities or large metropolitan areas.
• Data that was/is not readily accessible was used.
• Unpublished data sources were used.
• Subjectivity was endemic to most of the classification systems.
• Differentiating selection criteria varied within studies, often without explanation as to why.

This study counters the problems found in previous studies: it is inclusive of all employment categories; it is based on total employment, not labor force; it is inclusive of all metropolitan and micropolitan areas; it does not use unpublished data; all data is readily available and accessible; and the selection criteria are objective.

2.1. Units of analysis

The units of analysis were all the Metropolitan Areas (MA) and Micropolitan Areas (MI), as defined by the U.S. Bureau of the Census for 2020. A Metropolitan Area has a core country with an Urbanized Area of at least 50,000 people and a Micro Area with an urban area of at least 10,000 people but less than 50,000. Adjacent whole counties can be included if 25% of its workers are employed in the core county. There may or may not be additional Urban Clusters within the counties. Counties not including the primary urbanized areas must be linked to the central county through commuting patterns. The concept is that the Metro and Micro areas represent the economic city. They include physical areas and population which are not urban and thus include rural and agriculture. This causes a conundrum in the employment data discussed in the following section.

In 2020, there were 927 MAs and MIs excluding Puerto Rico: 384 Metropolitan Areas and 543 Micropolitan Areas. The smallest land area was the Metro Area of Vineyard, MA, at 103 square miles, and the largest was the Riverside-San Bernardino-Ontario, CA, Metro Area, at 27,300 square miles. The metropolitan and micropolitan area boundaries were downloaded from tlgdb_2021_a_us_nationgeo.gdb (accessed on 30 April 2022) [24]. The TIGER shapes were converted into points for use in mapping and display.

2.2. The data

Data on employment and occupation have evolved. There was a shift away from linking occupation and industry, and until 1940, it was possible to derive the type of business from occupational data as occupations were grouped based on industry. This was no longer true, starting with the 1950 Census [25]. Occupation is based upon the work performed and not the industry in which it is performed. Occupation was standardized into the Standard Occupation Classification (SOC) [25] in the 1970s. Industry was classified by the Standard Industrial Classification (SIC) code list. In the 1990s, this was revised and realigned with NAICS (North American Industrial Classification System), based on the production process of goods and services. This is a unifying feature of the NAICs classification. It is supply-based and production-oriented. [14] The groupings are producing units, not products or services. The bottom-line unit is a physical establishment. NAICs divides the economy into twenty sectors. There are 1,057 industries [26]. This resulted in a list of 236 detailed categories. For 2017, there were twenty major industry sectors. This is further regrouped into fourteen major categories for reporting in the American Community Survey. [25]

The breaking of the link between occupation and industry, plus the changes in classifying industry over time, renders it impossible to make comparisons over time.

The data used was employment by industry from the United States Bureau of the Census American Community Survey for 2021 [27]. Data on occupation was not used. There are 14 mutually exclusive categories [14]:

1. Military
2. Agriculture, forestry, Fishing and Hunting, and Mining (Sectors 11 and 21).
3. Construction. (Sector 23).
4. Manufacturing (Sectors 31–33).
5. Wholesale Trade (Sector 42)
6. Retail trade (Sectors 44–45).
7. Transportation and Warehousing, and Utilities (Sectors 48–49 and Sector 22).
8. Information (Sector 51)
9. Finance and Insurance, and Real Estate and rental and leasing (Sector 52 and 53).
10. Professional, Scientific, Management, Administrative, and Waste Management Services (Sectors 54, 55, and 56).
11. Educational Services, Health Care, and Social Assistance (Sector 61 and 62).
12. Arts, entertainment, Recreation, and Accommodation and Food Services (Sector 71 and 72).
13. Other Services, except Public Administration (Sector 81).
14. Public Administration (Sector 92).

See S2 Appendix NAICS Employment Categories for more descriptive details.

Only 895 of the Metro/micro areas had employment data reported in the 2001 American Community Survey. Only 513 had military employment reported. For this study, zeros were entered for military employment in areas with no reported military employment.

2.3. Methods

The methodological issue has been determining when employment in a category is significant, which is dominant, or what the cut-off point should be. Roy Kass reported in 1973 that there were five approaches: (1) based on absolute constraints; (2) based on the normal distribution; (3) based on the minimally necessary level of activity for a community’s viability; (4) based on the relative proportion of activity; and (5) based on a profile of the community [26]. Excluding studies based on subjective criteria, the studies based on objective criteria have used a variation of one of these three methods. Alexander notes that there are only three objective systems.

The first method classifies a city based on the three most extensive activities within a city. For example, in 2021, the New York Metropolitan Area’s most prominent employment category was Educational and Health Services (27.5%), followed by professional services (14.8%) and retail trade (9.6%).

The second set of methods uses positive differentials, usually from a national average. The first and most basic technique is subtracting the national average from the entity’s percentage. The results are different for the New York Metro Area. The top category now becomes professional services (a differential of 6.5%). Second becomes financial services (a differential of 4.4%), and third becomes educational services (a differential of 3.5%).

A second technique within the set of positive differential methods is formalizing the process using standard deviations. This was the technique used in the Nelson classification discussed above. New York City was classified as F2, two standard deviations for finance. The employment data used in this study were converted to z-scores. Using 2021 data, the most distinct category for the New York MA was information (z-score of 2.6). Next was financial services (z-score of 2.1). Professional services were third (z-score 1.9).

The problem with this technique is the use of the statistical mean. Extreme values, low or high, highly influence it. The distributions of the MAs within the employment categories were abnormal, so using statistics that require the normality assumption becomes problematic [7].

The third technique within this set of positive differential methodologies uses location quotients (LQ). These were used extensively in the classifications used by Duncan [8,23,28]. The location quotient calculates the metro area’s proportion of its total employment in each employment category by the same proportion for a larger area, in this case, national metro/micro total employment. For example, the New York metro area had 27.5 percent of its total employment in education and health services. For all Metro/micro areas, the proportion was 23.3 percent. Thus, the location quotient was 1.18. For information services, the LQ was 1.71. Ignoring the issue of cut-off points for now, New York had six categories of employment with Location quotients greater than one: information (1.71), FIRE (1.34), Professional (1.18); Educational and medical (1.18); Transportation and Warehousing (1.06); Wholesale Trade (1.03).

The accepted practice for interpretation is that if the LQ is one or less, the employment is classed as local or non-basic. Over one indicates a more significant share than the national [1,28] and is termed basic employment. Higher breakpoints have been used, including 1.25 and 1.5.

A third set of methodologies uses factor, discriminant, or cluster analysis. Roy Kass completed one such study in 1973 using cluster analysis of location quotients. Kass looked at the entire configuration of location quotients for a metropolitan area. The effort was to find those areas that had commonalities. The cluster analysis procedure produced twelve categories. Our example of New York was classified as retail and finance. In a 1977 study, Richard Wanner also employed factor and discriminant analyses to test two fundamental dimensions of large urban places – an industrial structure and a finance-commercial structure [8].

Tim Brennan also warned, “A danger is that the easy availability of computer algorithms may encourage thoughtless or inappropriate use of complex classification methods to produce a confusing mass of nonconverging and perhaps artificial empirical typologies.” [11]

This study used the location quotient to determine specialization, with the base for comparison being the total employment for all metropolitan and micropolitan areas, not the mean. The equation is:

To continue with the New York Metro Area example of information services:

For this example, the proportion of information services in the New York metro area was 1.7 times the national proportion. In contrast, the highest LQ for information services nationally was 3.5 for the San Jose-Sunnyvale-Santa Clara, CA Metro Area.

The selection criteria varied according to the data distribution for the industry category. The differentiating criteria were based on outliers to determine which activity stood out or was dominant. This is like using standard deviations but avoids the pitfalls of relying on the mean. Extractive, Military, and Construction criteria were set to outliers above the upper outer fence. The value for the upper outer fence is set to the third quartile plus three times the Interquartile Range for each of these employment categories. The criteria for the other eleven categories were set to outliers above the lower outer fence, which is the third quartile plus 1.5 times the interquartile range. These selection cut-off points are presented in Fig 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Location Quotient Descriptive Statistics for the Fourteen Employment Categories, 2021.

https://doi.org/10.1371/journal.pone.0334284.g002

For example, the selection cut-off for Extractive industries was a location quotient greater than 9.66. For Military employment, the cut-off criterion was a location quotient greater than 1.38. For Construction, the selection criterion was a location quotient greater than 2.6. For other industrial categories, the selection criterion was greater than the upper-lower fence. Examples include Manufacturing, set to 3.34; Retail Trade, set to 1.59; and Public Administration, set to 2.29. See Fig 2 for a list of all the selection criteria for the fourteen categories.

The Coefficient of Specialization (COS), also known as the Index of Dissimilarity or the Index of Divergence, was determined to get generalizations about a metro/micro area’s overall distribution of industries. Duncan [24] and Isard [30] discussed the COS. The COS “… measures the extent to which the distribution of employment by industry classes in the given region deviates from such distribution for the United States.” [29], though here, the total distribution was for all metro/micro areas. It shows the divergence of the metro area structure from the national structure. The higher the value, the greater the dissimilarity, or divergence, between the metro area and the national.

The coefficient is calculated for each metropolitan/micropolitan area for each industrial category by subtracting the area’s proportion from the total national proportion, summing either the negative or the positive differences for the area.

E = employment

i = industry category i

a = metro area

n = total national

C = number of categories

The software used was Microsoft Excel (with the add-ins StatistiXL [30] and AbleBits [31]), Microsoft Access, and ArcGIS Pro.

3.1. Classification schema

A general theme throughout the past research was the flow of material through the economy in stages. The overall categorical structure presented by Ergon Bergel in discussing Roderick McKenzie’s system forms the kernel of top-level categorization, or Class, used in this study after adjusting for the newer employment categorization and building on the NAICs classification and modifying it to fit the task of classifying metro areas. The classification schema used in this study was Military, Extractive, Manufacturing, Distributive, Trade, Services, Multiple, and Diversified. These eight classes are discussed in the following sections (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. The Functional Classification Schema.

https://doi.org/10.1371/journal.pone.0334284.t001

3.2. General

Fig 2 summarizes the descriptive statistics (mean, median, lowest, highest, first and third quartiles, and the interquartile range) for the location quotients for each industrial category. The military had the highest location quotient, recording a value of 45.3 for an equivalent local percentage of 38.1%. The next highest location quotient was 16.4 for the extractive industries. The manufacturing sector registered a median location quotient of 1.2, with the highest being 4.4 (local equivalent of 42.8%). Five industrial categories had zero as their lowest location quotient (military, extractive, information, FIRE, and professional services). The largest interquartile range (IQR) was for the extractive industries at 2.2, followed by manufacturing with an IQR of 1.03. The extractive industries had the largest median location quotient of 1.7, and Manufacturing had a median location quotient of 1.2. The lowest median location quotient (excluding the military which was zero) was 0.6 for the information sector.

The distributions of the metropolitan/micropolitan areas within the categories were far from normal, as can be seen graphically with the box-whisker plots (Fig 3). The distributions were either very skewed, had excess kurtosis, or both. A normal distribution would have a kurtosis value around three and zero skew. For example, the military category had a skew of 7.06 and a kurtosis of 65.2. This means there was a high peak and an extremely long tail. Public administration was another example, with a skew of 2.89 and a kurtosis of 18.15. Nine of the fourteen categories had excessive kurtosis with a value over three. All had a positive skew, and nine were highly skewed.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Distribution of Metro and Micro Areas for Each Employment Category, 2021.

https://doi.org/10.1371/journal.pone.0334284.g003

Some metro/micro areas reported no employment in some functions excluding the military sector. Fifteen metro areas had no employment reported in wholesale trade functions. Forty metro areas reported no employment in information functions. Only one area reported zero in FIRE. Only one reported zero in professional and only one in extraction functions.

3.3. Multiple specialization class

The observation that stands out the most dynamically is that not many areas have multiple functional specializations. As can be seen in Table 2, only forty-five areas out of the 927 Metropolitan and Micropolitan areas (4.9%) had two or more specializations. Only five (just over half a percent) had three specializations: Elk City, OK; Espanola, NM; Seattle, WA; Sterling, CO; and Washington, DC.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Metropolitan & Micropolitan Areas Which Have Multiple Specializations, 2021.

https://doi.org/10.1371/journal.pone.0334284.t002

The presence of Washington, DC, on this list is not surprising, as it is the U.S. capital. Public administration had a location quotient of 2.7, almost three times the national percentage in public administration. The military had a location quotient of 1.81, while not exceptionally high, and it was still over the upper outer fence cut-off (1.38) for military employment. Combined, these three categories totaled 36.82% of DC’s employment. Seattle, Washington, also was not a major surprise. Information industries (economic sector 51) include industries that produce and distribute information and cultural products or transmit and process data. Publishing and printing industries are found here, as are broadcasting, telecommunications, and data processing centers such as a GOOGLE Data Center [14], which had a location quotient of 1.8, with the upper inner fence cut-off of 1.6. The military industry also had a concentration with an LQ of 1.7. Professional services (economic sectors 54, 55, and 56, including accounting, mapping, architecture, engineering, consulting, veterinary, and legal services [14]) had an LQ of 1.41, while the upper inner fence was 1.3. A total of 22.54% of the metro areas’ employment was in these categories.

The other three on the list were intriguing. For Espanola, New Mexico, L & H (arts, entertainment, recreation, accommodation, and food services) had a location quotient of 1.79 (the cut-off was 1.7). This is economic sectors 71 and 72. Included are establishments that provide live performances, operate cultural facilities, operate casinos, food services, and operate establishments for travel lodging and recreational lodging [14]. Professional Services had a location quotient of 1.4. Information services had a location quotient of 1.7. These latter two categories are not surprising as this is the home of the Los Alamos National Laboratories. A total of 35.81% of Espanola’s employment was in these three categories.

For Elk City, Oklahoma, wholesale trade had a location quotient of 4.5, the highest LQ for wholesaling in the country. Next in line was TWU (transportation, warehousing, and utilities), with an LQ of 2.7. TWU is industrial sectors 22, 48, and 49, including establishments that provide transportation for passengers or cargo by any means, including pipelines, warehousing or storage, electricity, and water supply [14]. Third was L & H, with a location quotient of 2.13. These categories garnered 43.74% of the metro’s total employment. For Sterling, Colorado, public administration had a location quotient of 3.27. Sterling is the location of a state prison. L & H was next with a location quotient of 2.30. Information industries were third with an LQ of 2.08. A total of 38.29% of the metro’s total employment was in these three categories.

Manufacturing industries only showed up once for the forty-five areas with two or more specialties. This was Dumas, Texas, specializing in meat processing, with a manufacturing location quotient of 3.57. The most common industry was public administration, with fifteen areas having them as one of their two specialties. The second most frequent industry was information, with thirteen areas having it as one of their two specialties. The third most frequent industry was the military, with eleven areas. Tied for third were Leisure and Hospitality specializations. Interestingly, three areas had extractive industries (agriculture and mining) as one of their two specialties.

There appears to be no consistency in the pairings. The most frequent pairing was between information industries and professional service industries, which occurred together in six areas. The second most frequent pairing was the military and public administration, with five areas.

The forty-five areas of the Multiple Class had a median COS of 0.221; 22.1% of those employed would have to change employment categories to match the national distribution. The Class had a range of 0.3064 between the low value of 0.088 and the maximum of 0.394. The metro area with the highest COS was Dumas, Texas. The multi-Class metro area with the lowest value was Seattle, Washington.

Geographically, the areas with multiple specializations were widely dispersed across the country. One of the metro areas with three specialties was located on the East Coast, one on the West Coast, and three were in the central region: Colorado, New Mexico, and Oklahoma. The remaining thirty-eight areas with two functional specialties were scattered across the country. There was a hole in the East North Central and the Middle Atlantic regions. So, in some sense, the distribution was in a “crescent” shape. (Red in Fig 4.)

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Functional Classification of Metropolitan and Micropolitan Areas, 2021.

https://doi.org/10.1371/journal.pone.0334284.g004

3.4. Single specialty classes

3.4.7. Diversified class.

Any metro/micro area that did not exceed one of the cut-off points for any classification category was classified as diversified. This large class contained 584 metro/micro areas (65.3% of all metro/micro areas). (No color in Fig 4.)

The Diversified Class had a median COS of 0.144. Nashville, Tennessee, had the minimum COS (0.031), and Huron, South Dakota, had the maximum COS (0.298). The range was 0.2677. Being classified as diversified does not mean that a metro or micro area lacks a concentration of employment in any given industry. It simply was not high enough to breach the cutoff point. The diversified class was not a totally homogenized mass. For example, Celina, Ohio, had a manufacturing location quotient of 3.31, but it was below the cutoff of 3.34. Another example was Keene, New Hampshire, which had a wholesale trade location quotient of 2.04 but was below the cutoff point of 2.05. A third example was Odessa, Texas, which had an extractive location quotient of 9.25 but was below the cutoff of 9.66. Chicago, Illinois; Dallas, Texas; Houston, Texas; and Philadelphia, Pennsylvania were among the areas classified as diversified.

As shown above, the cutoff points are very sharp – the lines are in fact fuzzy. To examine the areas that are close to cutoff points — those areas that were “close but not quite there”, the following process was employed. For extractive and construction functions, areas with location quotients that were between the lower outer fence and the upper outer fence were selected. For all other functions, those areas that had a location quotient within ten percent of the lower fence were picked. The choice of ten percent was admittedly arbitrary. A total of 153 metro areas fell into this group.

There were thirteen Diversified Metro/Micro Areas that had an orientation to the Extractive Class functions, falling between the upper inner fence (LQ = 6.36) and the upper outer fence (LQ = 9.66). Some were just under the cutoff threshold point for selection as a functional specialty. Odessa, Texas, had an Extractive LQ of 9.25, just short of the threshold. Next was Alice, Texas, with an LQ of 8.90. Both were centers for oil and gas extraction. Third was Wenatchee, Washington, with an LQ of 8.72, specializing in agriculture, mainly apples. This group had a Coefficient of Specialization that ranged from a low of 0.12 in Fresno, California, to a high of 0.30 in Huron, South Dakota.

Geographically, the diversified metro/micro areas that had some sub-dominance in extractive functions were all west of the Mississippi River. The Pacific region had five, the West North Central region had five, and the West South Central region had four. This pattern matches that of the Metro/Micro Areas that were in the Extractive Class.

Retail trade functions showed a large number of metro areas within ten percent of the upper inner fence (LQ between 1.43 and 1.59). Twenty-four areas were identified as having a moderate dominance in retail trade. At the top of the list was Fayetteville-Springdale, Arkansas, with a LQ of 1.57, which is the home base for Walmart. Dyersburg, Tennessee, was next with a LQ of 1.56, and the third was Eau Clair, Wisconsin, with a LQ of 1.55. Geographically, seven were in the West South Central Region, and seven were in the East South Central Region. None were located in the Mountain or the Pacific Regions. Of this group, Plainville, Texas, had the highest Coefficient of Specialization (COS) of 0.255, while Rutland, Vermont, had the lowest at 0.12 (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Diversified Class Areas With a Sub-dominant Function.

https://doi.org/10.1371/journal.pone.0334284.t003

Twenty-three metro areas fell into this zone of possible moderate dominance in Productive functions (manufacturing, information, and construction) within a ten percent range (for manufacturing, between 3.006 and 3.34; for information, between 1.44 and 1.60, and for construction, between 1.93 and 2.60). For manufacturing, Celina, Ohio, tops the list with a LQ of 3.31, followed by Bennettsville, South Carolina, with a LQ of 3.24, and in third place is Manitowoc, Wisconsin, at 3.23. The north-south axis was also identified in this group, although it extended to the northwest, encompassing Iowa, Wisconsin, and Minnesota. Of the thirteen metro/micro areas, 54% were in the East North Central census region, with six being in Indiana, Ohio, and along the southern border of Michigan. Four metro areas had a sub-dominance in information, three of which were in the Middle Atlantic and New England Regions. Texas had the remaining area. There were seven areas with moderate dominance in the construction sector. Topping this list was Taos, New Mexico, with a LQ of 2.56. The second was Okeechobee, Florida, at 2:31. These metro areas were scattered throughout the county, though none were located in the Pacific Region. The Coefficient of Specialization (COS) for these areas ranged from a low of 0.08 in Burlington, Vermont, to a high of 0.30 in Storm Lake, Iowa.

Interestingly, twenty-four metro areas had possible multiple sub-dominant specializations. Two of the areas had three. Steamboat Springs, Colorado, had a dominance in construction, FIRE, and other services. Stephenville, Texas, had a dominance in agriculture, retail trade, and information. Pairings of functions showed no consistency. Geographically, the Pacific Region had six of these metro areas, the Mountain Region had six, and the West South Central Region had five. The Coefficient of Specialization (COS) ranged between a low of 0.09 in Santa Cruz, California, to a high of 0.20 in Greenville, Mississippi.

In all cases other than these twenty-four, if a Metro/Micro area’s economy contained an industrial type that was moderately dominant, there was only one dominant.