British employer census returns in new digital records 1851–81; consistency, non-response, and truncation – what this means for analysis

Abstract Newly available digital resources from the British census identify employers and their workforce size. However, there was a non-response rate of about 2.3% for smaller firms, rising to over 10% for firms over about 300 employees, and higher for the largest manufacturing firms. Non-responses are largely random except for different forms of business organization: significantly higher for corporates, and lower for unincorporated enterprises, but with no significant differences between partnerships and sole proprietors. Proprietor age is also significant. Non-response derives from defective census design and administration. Transcription truncations are also evaluated, which are higher for the largest firms, and vary by sector and position in the response string. Guidance to researchers on weighting and robust estimation strategies are presented for dealing with these limitations.