The Story of DIME: A Progress Report

Although there were many important antecedents, the story of GBF-DIME began in the summer of 1965, when the Census Advisory Committee on Small-Area-Data was established (Smith, 1967). In January 1966, the Census Small Area Data Advisory Committee began to develop plans for a 'case study,' and in June of that year, the Census Use Study (CUS) was established in New Haven, Connecticut; the New Haven office of the study was opening in September 1966 with a director, an assistant project director, and a staff of five people (Census, 1970). The year 1967 was busy at the CUS, which played a major role in a pre-test of mail-out, mail-back procedures to be used for the 1970 census. The test census in New Haven was conducted on April 1, 1967 (Census, 1970). The CUS also began "computer mapping experiments" in February 1967, and started "geographic base file research" in April. The programmers at CUS were struggling with the inefficiency of inherent redundancies in the conversion of analog maps into numerically encoded renderings. In effect, each street intersection in a normal US. rectangular street grid was being digitized 8 times (Cooke, in press).

Donald Cooke, who at the time was a programmer on the CUS staff, reports that the problem was overcome based on principles of map topology presented to the CUS staff by Census mathematician James Corbett (Cooke, in press). The encoding scheme later known as DIME (Dual Independent Map Encoding) was developed. The key idea was to number the nodes (street intersections) and the areas (typically blocks). By encoding street segments in terms of the areas to the left and right of them as well as the nodes that they connected (to node, from node), the topology was encoded with redundancy that allowed automated checking for consistency. This redundant coding of nodes and areas adjacent to each 1-dimensional object is at the core of the 'chains' or 'arcs' structure underlying modern vector GIS data models such as DLG (USGS Digital Line Graphs), SDTS (Spatial Data Transfer Standard), and the polygon layers in commercial systems such as ARC/INFO.

In the summer of 1967, the significance of the innovation was more practical: it supported efficient digitization and error removal, and laid the ground work for choropleth mapping of census results. George Farnsworth of the CUS christened the new process "DIME" (Dual Independent Map Encoding) in August 1967, Cooke and Maxfield wrote a paper about the encoding, and they were "squeezed" into Robert Barraclough's session on computer mapping at the 1967 URISA meeting the following month (Cooke and Maxfield, 1967; Cooke, in press). While not on the CUS staff, Barraclough was at the time part of a technical advisory committee to CUS. Thus, this key innovation in the history of GIS went from the spark of invention to academic publication in a period of 3-4 months! GBF-DIME files were digitized for all US cities during the 1970s, and were a key component of the current TIGER system that in turn is a critical part of the National Spatial Data Infrastructure (NSDI). The prominence of the DIME encoding in the history of GIS, as seen by academics, has almost certainly been enhanced by its very early publication.

The story presented above seems simple and straightforward. However, there are some very interesting developments surrounding the Census Use Study. One centers around a parallel technical development, Robert Dial's "Street Address Conversion System" (SACS), and the other involves the promotion of commercial uses of DIME and small-area census data that contributed to the development of the US Geodemographics industry.

SACS: Street Address Conversion System

In 1964, Robert Dial completed a master's research project that was published in the form of a research report entitled "Street Address Conversion System" (Dial, 1964). This highly innovative research was conducted under the direction of Professor Edgar M. Horwood of the University of Washington, a founder of URISA in 1963. Dial's SACS system was similar in intent to DIME, but different in a number of ways. An important SACS innovation, not found in DIME, was the use of a grid for indexing nodes. Early in 1965, Dial described SACS to William Fay, head of the Geography Division of the US Census, at a lunch arranged by Alan Voorhees (Dial, personal communication, 1996). Dial reports that the Census person appeared very impressed with the approach (Dial, personal communication, 1996). Recall that this was some 3-6 months before the Census established the advisory committee on Small-Area-Data.

Donald Cooke reports that he and William Maxfield, the Census Use Study programmers who developed DIME, were not aware of Dial's SACS approach in 1967 (Cooke, in press). But this does not mean that no one at the Bureau of the Census knew about it. The Census Use Study in New Haven was operating rather independently from the Geography Division of the Census in Suitland, Maryland (Cooke, in press); William Fay headed the Geography Division until 1971 (Cooke, in press), and had championed ACG [Address Coding Guide] and resisted DIME (Cooke, in press). In 1967, Calkins (1967) presented a paper to the 46th Annual Meeting of the Highway Research Board, that reported on the application of Dial's SACS scheme to Ottawa, Canada, and cited Dial's 1964 report. Several people from the US Bureau of the Census or otherwise associated with the Census Use Study, including Robert Barraclough, Morris Hansen, and Robert Voight, also presented papers at that same Highway Research Board meeting. Two researchable questions are evident here. One would be a technical comparison of SACS and DIME in formal terms; the other is to interview key individuals from the CUS, the Geography Division of the Census, and elsewhere in the URISA and transportation communities, to determine sequences of events during 1967, and who knew what, when, about SACS and DIME.

Data Uses in the Private Sector

The evolution of the Census Use Study, and particularly the push to promote the use of Census data in the private sector also deserves attention. The New Haven office of the CUS closed in June of 1969, except for Health Information System activities (Census, 1970), and the Southern California Regional Information Study (SCRIS) was established by CUS the following month in Los Angeles (Census, 1973). The CUS organized a number of workshops, and of these, the one held at the Asilomar conference center in Pacific Grove, California, on October 4, 1973, is especially interesting. This workshop was entitled "Data Uses in the Private Sector," and although many of the speakers were from the public sector, other participants were largely from the private (Census, 1974). Vincent Barabba, at the time recently appointed Director of the Census Bureau by President Nixon, gave a talk that stressed commercial uses of census data. It was around this time that some of the major geodemographics firms, such as Claritas, began providing marketing services based on geodemographic analysis. Research should investigate the connections between the census data workshops and the early days of the geodemographics industry.