The year 2025 marks the return of the Census. Also known as the 'Population and Housing Census', the Census is the largest national statistical survey conducted by the Statistics Korea once every five years, targeting the entire population. The statistics obtained from the Census provide fundamental information about our lives, such as population, households, and housing. However, they also serve as the most 'important' record for understanding the social and economic conditions of the era. It has already been 100 years since the Census was first conducted. What does the history of South Korea, as shown by the Census over the past century, look like? We reflect on this through the '100 years of Census'.
A local government still hangs a faded sign saying 'statistical workroom'. It is a remnant of the past when statistics were manually compiled by hand. Similarly, the early Censuses were analyzed through 'manual aggregation'. Now, 100 years later, data processing has become automated, and with the introduction of artificial intelligence (AI), subjective responses can now be classified automatically.
◇ manual work→punch cards→computers→OMR→PC→ICR (optical character recognition)→AI
After the first Census in 1925, which was entirely conducted through aggregate collection, the 1930 Census introduced the Punch Card System (PCS). This system, which records selected responses by punching holes in paper that can be read by machines, represents the initial format of data storage. The Punch Card System slightly reduced the computation time for aggregate collections.
Then, in 1966, the Census (postponed for a year due to budget constraints) underwent a groundbreaking change with the introduction of 'computers'. The 'IBM 1401', known as Korea's first computer, was brought in to process the population census data.
The fact that this computer was operated in a meeting attended by President Park Chung-hee shows how monumental the event was. On June 24, 1967, the Dong-A Ilbo reported that 'to fully analyze the results of the 1966 population survey would require 450 personnel and 14 and a half years, but with the introduction of the computer, this time could be reduced to a year and a half.'
In 1990, Optical Mark Recognition (OMR) machines were introduced. OMR, commonly used in the college entrance exam, was employed for data entry by surveyors receiving the questionnaires at the time. However, initially, the reading period was long; it took 17 months for the final results to be released. By 1995, OMR reading times were halved to 8.7 months, allowing for the first time to publish a report within two years of conducting the Census.
In 2000, data entry was carried out through personal computers (PCs) at 18 locations nationwide, and by 2005, the Census field investigation could be conducted via individual 'internet' access. The participation rate was low at 0.9% in the first year, but in 2010, 47.9% responded through internet surveys. Consequently, from the mid-2000s, the survey results could be announced the following year, increasing their relevance.
In 2010, the Intelligent Character Recognition (ICR) method was also introduced, allowing computers to recognize handwritten characters.
In 2020, due to the COVID-19 pandemic, 'non-contact' methods became crucial, leading to the first implementation of the 'paperless population and housing census.' Respondents could participate via internet surveys through PC or mobile, and when a field visit was necessary, instead of paper questionnaires, surveyors entered data on tablets.
◇ total survey→1966 comprehensive + sample survey→2015 registration census
Not only have data processing technologies evolved, but the 'survey methods' have also progressed. The early Census, which conducted total surveys on all items, reached a turning point in 1966 by mixing 'total' and 'sample survey' techniques. Simple and essential items were fully surveyed, while more in-depth surveys on items like fertility and economic activity status were conducted only on 10% of the survey units. Since then, the mixed method of total and sample surveys has continued, with the sample survey ratio now at 20% of the population.
Until 2010, all items in the comprehensive survey were collected through on-site investigations. However, the introduction of the 'registration census' in 2015 eliminated this cumbersome process. Thanks to administrative data, it became possible to gather sufficient information. As a result, items collected through the registration census are now gathered annually, instead of every five years. Respondents for in-depth items, which can be determined through administrative data, only need to respond if selected as sample households.
Starting this year's Census, the Statistics Korea plans to actively utilize AI technology. For example, the questions concerning industry and occupation surveys are subjective, but now AI can analyze these responses and automatically assign standard classification codes. The 24-hour inquiry response system will evolve as well. In addition to human operators, AI voice and text consultations will also be available. Furthermore, the user interface (UI) and user experience (UX) of mobile response screens will be improved to be more user-friendly.
The continued advancement of Census technology ultimately aims to reduce the inconvenience for respondents and enhance the accuracy of surveys. However, no matter how much the technology is upgraded, growing concerns about personal information exposure are challenges that the Census must overcome.
Kim Jin-woo, director of the Statistics Korea Survey Management Bureau, noted, 'In the past, every neighbor participated without exception, but nowadays it seems that only one in five households participates, which creates reluctance.' He emphasized that national statistics like the population and housing census are still critical in establishing national policies. He expressed hope that those selected as sample households would participate with pride as representatives of a '50 million population.'
According to Article 33 of the Statistics Act, the information gathered is strictly confidential and will not be disclosed to others. Information obtained during the survey will only be used for statistical production purposes and will not be provided to others, so participants can rest assured. This year, the population and housing census will be conducted from Oct. 25 to Nov. 18.
☞Reference materials
Kim Min-kyeong, former vice administrator of Statistics Korea, 2018, 'Korea's population and housing census.'
Jo Eun-joo, 2014, 'Demographic statistics and nation-building: focusing on the 1960 and 1966 population censuses,' Korean Sociology, Vol. 48, No. 5.