India, June 18 -- It's a much delayed Census, and changes are afoot. On Monday, the government notified the 2027 Indian Census, which will be the first census exercise in India in 16 years. India has seen significant economic, political, and technological transformation in this time period. There is also a lot riding politically on the Census. It will likely be the basis for the controversial National Register of Citizens (NRC), which can be used to determine citizenship in India, should the government decided to extend the National Population Registry to a NRC; the next round of delimitation (drawing constituency boundaries); and caste reservations. This iteration of the Census will also include a "caste census" - enumerating Indians by caste for the first time since 1931, which will certainly impact the allotment of reservations for other backward classes (OBCs). With so much riding on the Census, a curious little addition to the census procedure, with no details, was included in the notification. For the first time ever, India will allow "self-enumeration" which will allow citizens to fill out Census data without a visit from a census enumerator. Let us take a closer look at the challenges and implications of self-enumeration. Many developed countries, such as the US, essentially use self-enumeration in the Census - as door-to-door canvassing is not operationally or fiscally feasible. Given India's advancement in biometric identification, greater mobile phone connectivity, and the difficulty of interviewing populations on the move such as commuters and migrants, it makes sense for the government to experiment with this idea. Beyond operational ease, self-enumeration can offer many advantages for procuring high quality data. It is no secret that demographics in many parts of India are highly sensitive matters. To take one example, speaking Bengali as a first language in the state of Assam has become politicised. Imagine how a Bengali speaker may respond to the language questions in the census if directly speaking to an Assamese-speaking enumerator. The interaction may cause the speaker to lie about their first language, or the enumerator may intimidate the respondent to generate incorrect data. Similar claims can be made about declaring oneself as Muslim where there is a threat of religious intimidation or being a Hindi-speaker in Karnataka. In principle, if privacy protocols are sufficiently strong, respondents may more honestly report their demographics and this could yield some surprising (but more correct) results. In this manner, the push to self-enumeration may help reduce "surveyor bias." At the same time, there can be genuine concerns about data quality with self-enumeration, from a greater likelihood of data entry error to purposeful misrepresentation. The Census data is complicated, and enumerators are given strict guidelines about how to code various definition (the definition of the household). How will these guidelines genuinely be communicated to the citizen, and will people be able to apply them systematically? Are we certain that each person that desires to enter data through self-enumeration will have the functional literacy to do so? But beyond entry error, there are genuine concerns of wilful misrepresentation. If we are concerned that the Census can only be filled by persons with a certain level of functional literacy, then it is equally true that those that do not reach that level will be obliged to look to others, such as "agents," to fill out the information. Given that this data will now be tied to a caste census and to a national register, will there be a secondary market that develops to fill out the information in the "right way"? For instance, will better organised caste groups get "better enumeration" through self-enumeration? Data fidelity is a major concern, even in the US. After the information is collected, the US runs a Post-Enumeration Survey to characterise the bias and error in self-enumeration. Is this something to which the government is committed? Will there be sufficient coordination in the system to know who has engaged self-enumeration and who needs a visit from a Census enumerator? In other words, when there are multiple ways of entering in the system, there are genuine concerns of redundancy but also miscoordination that misses out on Indians. Perhaps this initial run a self-enumeration will see a low adoption rate. But if the government is committed to moving to a system of self-enumeration for a significant share of the population, these are all issues that will have to be addressed. We all expect some growing pains. The biggest challenges will occur if self-enumeration throws up some surprising results, like higher than expected number of Bengali speakers in Assam. Will the data be seen as more accurate or the subject of manipulation? How will the validity of such surprising data be adjudicated? Given the operational costs of the Census, it was only a matter of time before the government looked to self-enumeration. As the government builds out the guidelines on self-enumeration in future notifications, it will need to think through these concerns to guarantee that the data remain high quality....