Sampling Strategy for the Telsur/Atlas Project

Sharon Ash

The sampling strategy for the Telephone Survey/Phonological Atlas project was designed with the goal of representing the largest possible population, with special attention to those speakers who are expected to be the most advanced in processes of linguistic change. It has been well established that phonological change is usually most advanced in urban centers; thus the first tier of communities to be sampled consists of places with the greatest concentration of population. Each community is selected as the focal point of an area, and the areas are determined so as to cover all the territory of English-speaking North America. Three defining terms are involved: Zones of Influence, Central Cities, and Urbanized Areas. The selection of places to be sampled involves intersecting characteristics of the three levels, as will be explained below. The terms will first be defined, and then the selection criteria that produce the overall sampling plan will be described.

1. Zone of Influence

A Zone of Influence consists of a number of counties. It is derived from the 1992 County Penetration Reports of the Audit Bureau of Circulations (ABC). ABC audits data from member organizations on the circulation of newspapers and other publications. For every county with at least 100 households, the County Penetration Report lists the name of each member newspaper, gives its circulation, and indicates whether it is a daily or weekly and morning or evening publication. A ZI, defined for the Telsur/Atlas project, is determined by Central Cities (see below). A county belongs to the ZI of a given Central City if, in that county, the circulation of the newspaper(s) from that city is greater than the circulation of the newspaper(s) from any other city that has been designated a Central City for the purposes of the research project.

Once the Central Cities have been selected, it is in theory possible to assign every county to a ZI. In practice this is not true, because some counties have fewer than 100 households and so are not listed in the County Penetration Reports. In most cases, such counties can confidently be assigned to a ZI on the basis of the ZI assignment of surrounding counties. In a few cases, the assignment of a given county could arguably be made to either of two ZIs. In those instances, the assignment was made on the basis of considerations such as proximity to the Central City.

2. Central City

This term is used in two senses. First, it is used as a synonym for the U.S. Census Bureau's definition of a Central Place as the defining feature of larger census units, including the Standard Metropolitan Statistical Area (SMSA) and the Urbanized Area (see below). The second sense is defined for the Telsur/Atlas project: a Central City is the central place of a Zone of Influence. As in the Census Bureau definition, a Central City may actually consist of more than one city: examples are Minneapolis-St. Paul, MN and the Quad Cities on the Mississippi River (Moline and Rock Island in Illinois and Davenport and Bettendorf in Iowa.) The basic criterion for the selection of a Central City of a ZI is that it be a place for which the Urbanized Area (see below) has a population of at least 200,000 according to the 1990 census. Due to low populations in some areas, it was necessary to designate a number of cities smaller than this limit as Central Cities, such as Burlington, VT, Roanoke, VA, and Boise, ID. Three of the Central Cities are even smaller than the threshhold of 50,000 which is used by the Census Bureau as a criterion for status as the Central Place of an Urbanized Area; they were assigned the designation of Central Cities for the same reason as the other Central Cities with a population under 200,000: to provide well-motivated geographic coverage. The status of such towns as regional centers is demonstrated by the existence of a local newspaper that has wide circulation in the area. The three Central Cities which are not UAs are Minot, ND, Aberdeen, SD, and Rutland, VT. Thus a Central City serves as the defining place of a Zone of Influence, and at the same time it is the Central Place of an Urbanized Area. Figure 1 shows the Zones of Influence, with a star locating the Central City or Cities of the Zone. The Zones are labeled with 2- or 3-letter abbreviations for which the key is given in Table 1.

3. Urbanized Area

This term is defined by the U.S. Census Bureau in order to provide a better separation of urban and rural population than is given by the SMSA, which takes the county as its building block. It consists of a central city or cities and the surrounding densely settled territory. By definition, it has a population of at least 50,000. The densely settled surrounding area consists of contiguous incorporated or census designated places having either a population of 2,500 or more, a population density of 1,000 persons per square mile, a closely settled area containing a minimum of 50% of the population, or a cluster of at least 100 housing units. Further details on the definition of an Urbanized Area may be found in the Census Reports. The composition of each Urbanized Area is defined by maps in the series of census reports 1990 CPH-2: Population and Housing Unit Counts. In the design of the Telsur/Atlas sample, the Urbanized Area is taken to be a conservative estimate of the territory of the speech community of the corresponding Central City. If a speaker is a native of any place within the Urbanized Area of a Central City, s/he is taken to be linguistically representative of the Central City's speech community. The areal extent of the UAs as mapped by the Census Bureau is quite restricted, which allows us to be confident that this is a valid sampling decision.

The Central Cities selected to define ZIs are further divided into four types by population of the corresponding UA and by area of the ZI, as follows:

p1 UA population > 1 million
p2 UA pop. > 200,000, non-restricted (area > 5,000 square miles)
p3 UA pop. > 200,000, restricted (area < 5,000 square miles)
p4 UA pop. < 200,000

These four levels are used to differentiate the amount of sampling to be done in smaller cities within each ZI. At the level of the Central Cities, the only difference in sampling is between the p1 cities and all others: in p1 cities, at least four speakers are to be interviewed, while in all others, at least two are to be interviewed. Furthermore, in every city, at least one speaker must be a woman between the ages of 20 and 40.

Table 1 lists the 145 Central Cities that have been selected for sampling and gives the corresponding ZI and UA populations. The figures show that 54% of the population of the United States lives in the 145 Urbanized Areas (or smaller cities) that have been selected for sampling. Thirty-three of the UAs have a population over one million, and 112 have a UA population under one million. Thus the total minimum number of speakers to be represented in the completed national sample of the United States is 356 speakers. (A similar sample, consisting of about 40 speakers, will be designed for Canada.) A sizable number of speakers from smaller towns have already been interviewed as part of an earlier research project, and it is often necessary to contact additional speakers in a locality in order to reach a woman in the desired age range. Therefore, the total number of speakers to be interviewed will be considerably greater than 356. To date, approximately 550 speakers have been interviewed, and about 30 more are needed to complete the sample in the United States. The additional speakers add greatly to the depth and richness of the data.