by Hillary Vipond (London School of Economics)
This blog is based on research made possible by the Carnevali Small Research Grant Scheme, funded by the Economic History Society.
Census data for nineteenth Britain contains additional information on occupation which has been all but hermetically sealed from analysis by the constraints of processing ‘big’ data. For the last two hundred years the data for occupation has been tracked at industry level: this provides approximately 800 categories of occupation. However, a sub-industry level of data, providing approximately 15,000 different categories of occupation, is concealed within the textual descriptions of occupation recorded when the census data was originally collected. My project processes this text to construct the sub-industry, task-level, variable for occupation.
Every British census between 1851 and 1911 required individuals to describe their occupation. William Farr, superintendent of statistics at the time, indicated that the description of occupation was meant to reflect five key aspects of work: ‘skill, talent, or intelligence; tools, instruments, machinery or structures; materials; processes; products’, and indeed these elements of the work are often present in the data. The responses given by householders ranged from terse one-word summaries through to detailed descriptions. See, for example, the variable for ‘Occupation’ (Table 1).
|Table 1: Example data (bootmakers only)
|Industry Code||Occupation (Original response as written in the Census data records)||New Variable: “Task”|
|663 (Bootmaker)||BOOT & SHOE MAKER (MASTER EMPLOYING 4 MEN 4 WOMEN & 5 APPS)||Maker|
|663 (Bootmaker)||BOOT AND SHOE BINDER||Binder|
|663 (Bootmaker)||BOOT & SHOE RIVETTER||Rivetter|
Note: Example of Individual-level observations, first two columns as in ICEM digitized data, third column showing the newly constructed “task” variable.
In 1851, following the collection of the census that year, the General Registers Office received hundreds of thousands of these unique character strings describing occupation. This body immediately realized that analysing the data at such a granular level was not feasible, and set about creating an industry level categorization scheme. This consisted of 800 different industries, and clerks were hired to assign individual records to the appropriate industry.
Social science research based on the nineteenth century British census data has predominantly been conducted at the industry level of occupation. When the Integrated Census Microdata project (ICEM) digitized the English Census data between 1851 and 1911, it retained the industry level classification of occupation. However, it also digitized the character strings which were the original responses to the question on occupation.
My project extracts the sub-industry ‘task’ level information from these character strings to construct a more granular categorization. The process used identifies the set of tasks which account for at least 95 per cent of the individuals in each industry. The data reveals the evolution of the occupational structure within an industry. For an example, see Figure 1, below, which tracks tasks for English bootmakers, watchmakers , laundresses, and printers.
Figure 1. Tasks by industry, 1851-1911
This more granular data provides new insights. For example, it uncovers the impact of the adoption of new technology on occupational structure, which is obscured when viewed solely at the industry level, as shown in my paper on the mechanization of the English bootmaking industry: link. In this industry, task level occupational data show that there was a sharp decline in the number of ‘binders’, responsible for hand-sewing the upper portion of the shoes together, while new jobs were created for ‘sewing machinists, who performed the same work but with the assistance of machines. The timing of this transition coincides with the adoption of sewing machines in the boot making industry. In the watchmaking industry there is little evidence of mechanization, but the general decline of the industry seems to coincide with an increase in the number of watch and clock dealers. In the laundry-cleaning industry some mechanization is apparent, but most of the work was undertaken by washerwoman using manglers. Conversely, the printing industry grew rapidly, and utilised mechanization considerably.
My preliminary results indicate that specialization within an industry may have increased in tandem with mechanization. They also provide evidence that some industries, or the tasks which comprised them, became more spatially concentrated. This has implications for the changing geographies of tasks and skills in England, and, potentially, access to opportunity, particularly in the context of the nascent North-South divide. Finally, the new data reveals the pace and geographical diffusion of technology adoption in Britain; this is potentially valuable in its own right as there is little accounting of the capital stock during the second industrial revolution.
To contact the author: