27 | Census under Pandemic and Protest: The 1921 Census of India

India in 1921 remained overwhelmingly rural. A town was defined, as in the two previous censuses, as any municipality, cantonment, civil lines area, or any “continuous collection of houses” of 5,000 or more persons that a provincial superintendent chose to treat as a town. 

Of the population for which urban data were tabulated, only 10.2 percent or  a little over 32 million people which were spread across 2,313 towns and cities, were classed as urban. The report clearly highlighted the comparison, noting, “as compared with 23 percent in the most urbanized unit of the Indian Empire [Bombay Presidency], the proportion of the population of England classed as urban is 79 percent, and of France 44 percent.”

One striking finding concerned the sex composition of India's largest commercial and industrial towns. Calcutta's population had males outnumbering females by distinctly over two to one, with only 374 married females per 1,000 married males. It reflected a predominance of male migrant labour which the report noted as “leading an unnatural existence, missing the comforts of home life.” 

Rangoon showed a similarly skewed ratio of 444 women per 1,000 men, against the resident Burmese town of Mandalay, which counted 915. This was the evidence that the imbalance was a function of migrant industrial labour rather than of urban life as such.

Table: FEMALES PER 1,000 MALES IN SELECTED CITIES, 1921

By contrast, the average rural sex ratio in most provinces sat much closer to parity, underscoring that the imbalance was specific to industrial and commercial towns that drew in male labour from outside — jute mill hands in Bengal, dock workers, mill labour in Bombay — while their families remained in the village.

The contrast of Villages

The chapter on the rural and urban populations closes with a self-critical discussion of what a village even meant statistically. The term covered, province to province, anything from a genuine settlement to a purely administrative revenue unit with no relationship to where people live.

In Sind, one revenue village in the Karachi district covered 1,806 square miles and 3,572 occupied houses but was recorded as containing only two villages. In Bihar's Monghyr district, the size of a revenue village (mauza) ranged from under one acre to 72,000 acres. Marten concluded that “there is no single homogeneous unit which can be described as a village,” and that the cross-province comparisons of rural distribution on this basis were of limited value.

The 1921 census recorded birthplace by asking every person to name their district of birth and, if outside the province, their native province, state, or country. Only 603,526 persons were recorded as born outside India altogether. About four-fifths of them were from other Asian countries (Nepal, Afghanistan, China, Ceylon, Arabia chiefly), the rest mostly from Britain and Europe.

India, in turn, was estimated to send out roughly 1.7 million people to other countries. The international movement in either direction was practically negligible at India's scale.

Movement within India was similarly modest. About 30 million people or 10 percent of the population, were enumerated outside their district of birth, and of these some 20 million had moved no further than a neighbouring district. The report attributed this to “economic and social causes, and the immobility of an agricultural population rooted to the ground, fenced in by caste, language and social customs.”

Five kinds of migration

The internal movement was organized into these five categories in the report:

1. Casual movement between neighbouring villages (marriage, family, minor business), 

2. Temporary labour migration to canals and railways, or pilgrimage and marriage travel, 

3. Periodic migration tied to harvest labour, 

4. Semi-permanent migration, where workers earned a living away from home while remitting earnings back, and 

5. Permanent settlement migration, as in the Punjab canal colonies or Assam tea gardens. 

Marriage customs themselves generated much casual movement, since a bride often married outside her village and returned to her parents' home to give birth, so birthplace statistics for women partly reflect marriage patterns rather than economic migration.

Of all the categories in the schedule, age was treated with the most open scepticism about the numbers. Marten traced it to a general disinterest in exact age and a residual unease about disclosing it at all. Edye's account from the United Provinces is candid about this:

The ordinary educated Indian has very vague ideas about his own age. The uneducated Indian has practically no ideas at all... During the period when the staff was being trained, I had my own age guessed by hundreds of supervisors and enumerators, and the estimates were seldom within five years of the truth, and varied between 16 and 60.

— W. Edye, United Provinces Superintendent

This showed up statistically as heavy ‘digit preference’,clustering of reported ages on multiples of five and, especially, of ten. In a specially tabulated group of 100,000 males in Bihar and Orissa, about 25 percent of ages were recorded ending in 0, and a further 18 percent ending in 5. 

The census utilized corrective techniques, including an “Index of Concentration” borrowed from United States Census Bureau practice, to quantify and adjust for this rounding bias.

Infant mortality

The report situated Indian infant mortality within an international comparison where death rates for infants under one year had fallen steadily in most countries since 1881 — by 24.6 percent in France, 15.8 percent in England, as much as 37.6 percent in Australia. 

In British India, about one-fifth of children died before their first birthday, attributed to biological factors (age of the mother, spacing of pregnancies) and economic factors (poverty, sanitation, malnutrition). Rates were highest in the United Provinces and Central Provinces, lower in Madras. Bombay and Calcutta had distinctly high rates, though the report cautioned that defective birth-reporting in cities may exaggerate the apparent rate.

Sex Ratio

India's sex ratio stood at 955 in 1921, continuing a slow decline from 962 in 1901. Going against the prevalent theory, the report devoted much energy to rebutting a charge occasionally made by critics, that this male excess reflected systematic under-counting of women rather than a genuine demographic fact. 

Marten retorted that the shortfall does not appear preferentially among communities where women are secluded, and the ratio worsened, not improved, as enumeration grew more accurate over successive censuses,which is the opposite of what an under-counting theory would predict.

The report instead locates the cause in the differential treatment of the sexes across the life course, quoting at length from Sir Edward Gait's 1911 analysis:

In India... sons are earnestly longed for while daughters are not wanted... Sometimes the prejudice against daughters is so strong that abortion is resorted to when the midwife predicts the birth of a girl. Formerly female infants were frequently killed as soon as they were born, and even now they are very commonly neglected to a greater or less extent... To the evils of early child-bearing must be added unskilful midwifery; and the combined result is an excessive mortality amongst young mothers.

— Sir Edward Gait, Chapter VI

(Maps showing proportions of sexes, actual vs natural. Source: Marten, 1924)

Marten adds that nothing in the specific circumstances of 1921 explains the further fall in the female ratio during this decade; if anything, he treats the consistency of the decline as itself an argument against any theory of episodic under-recording, and as evidence of structural causes operating steadily across decades.

The census classified every person, whether infant, child, or grown up, into the categories of unmarried, married, or widowed . The report flagged a distortion built into these categories for Hindu and Jain communities, because of child marriage. Many recorded as married were still living in their parents' homes awaiting a second ceremony, typically at puberty, before actually taking up married life. 

A decade of rising marriage age

Despite this caveat, the chapter's most notable finding is that of a measurable shift toward later marriage. This direction was the same across nearly every province. In Bengal, the proportion of boys aged 10–15 left unmarried rose from 826 to 868 per thousand between 1891 and 1921, and for 15–20 from 594 to 665. Mr Thompson, the Bengal Superintendent, offered a balanced explanation:

This may have been due in part to the fact that the last few years were hard times, and the greater difficulty of scraping together the money necessary to be spent... may have caused some marriages to be postponed, but the tendency of public opinion to favour later marriages has been a very significant one.

— Mr Thompson, Bengal Superintendent, Chapter VII

The pattern varied by caste. In Bengal, the Brahman and Baidya castes led the trend, followed by the Bengali merchant (Nabasakh) class. The Chasi Kaibartta cultivating caste raised the marriage age for boys but continued to marry girls young. Darbhanga in Bihar remained the “home of infant marriage”, with 64 boys and 103 girls per 1,000 married between ages 0 and 5. In the United Provinces, Edye attributed postponement among higher castes not to any shortage of marriageable women but to the extended schooling that was delaying marriage.

State legislation and widowhood

Two princely states had legislated directly on the question of child marriage. Baroda's Infant Marriage Prevention Law of 1904 set minimum ages of twelve for girls and sixteen for boys, with a lower exception carved out for the Kadwa Kanbi community's mass marriage ceremonies. The report noted stricter enforcement this decade, with rising convictions. Mysore had comparable legislation. 

On widowhood, the influenza epidemic left a visible mark, because mortality struck married adults heavily without killing both spouses equally. It mechanically transferred many survivors into the widowed category.

The report acknowledges that the statistics on infirmities are among the most difficult to collect accurately due to factors such as imperfect diagnosis, unintentional omissions, and the intentional concealment of afflictions by families fearing social stigma. The total number of recorded infirm persons in India saw a continuous decline between 1881 and 1901, largely due to improved medical knowledge and actual decreases in prevalence (such as cataract cures for blindness). 

However, between 1911 and 1921, there was a slight increase in recorded cases, which administrators attributed to more thorough recording and tabulation rather than a rise in the conditions themselves. This decade was also marked by high mortality among the infirm during the 1918 influenza pandemic and subsequent famines, as these individuals often lacked the resources or physical strength to survive such crises. 

The report noted a much lower proportion of insanity in India compared to Europe, though it questions whether this reflects a true difference in "mental strain" or simply the difficulty of diagnosing mental derangements through a general census. Deaf-mutism was defined as a congenital defect. A change in the census instructions in 1921 which omitted the phrase ‘from birth’, led to the erroneous inclusion of some elderly people with senile deafness, which skewed the data for older age groups.

Blindness was the easiest infirmity to diagnose and carried the least social stigma, making its figures the most reliable. It observed that women suffered more frequently than men, a trend attributed to their prolonged exposure to smoky fires while cooking in ill-ventilated houses. 

Statistics for leprosy were considered heavily underestimated, with experts suggesting the actual number of cases could be five times higher than the census figures. The disease was frequently concealed due to shame, and its prevalence was often linked to hygienic conditions and specific regional environments, such as the Naga Hills or Burma. 

In almost all categories except blindness, males were recorded in significantly higher numbers than females. Marten concluded this was not a biological reality but rather evidence of systematic concealment of female afflictions within the household. 

An inquiry into the Punjab region specifically examined whether consanguineous marriages (common among some Muslim communities) led to higher rates of infirmities. The analysis provisionally concluded that there was no clear evidence that such marriages were more productive of insanity, blindness, or leprosy than the practices of the Hindu community

The 1921 Census of India marked a significant shift toward standardized measurement by utilizing the "write a letter to a friend and read the reply" test adopted in 1911. This allowed for a genuine comparison across the decade, revealing that overall literacy rose from 59 to 82 per 1,000 persons, a 39% increase. 

Beyond simple counting, the 1921 report introduced sophisticated analytical tools, such as "cost-per-literate" calculations in the United Provinces and demographic forecasting models in the Punjab to project future educational needs. Despite these administrative advancements, the data highlighted a stark reality that nine out of ten Indians remained entirely illiterate by the census’s exacting standard. 

Regional and social disparities remained remarkably durable throughout the decade.

Table 1: Literacy per 1,000 persons by age group and province/state, Census of India 1921

Burma continued to lead the empire in literacy due to its traditional monastic schools, while the princely state of Kashmir remained at the bottom. The social hierarchy of literacy also persisted, with Zoroastrians (Parsis) maintaining a significant lead over other religious groups and scribal castes like the Baidya and Kayastha dominating the educational landscape. 

In contrast, depressed castes like the Chamar and Musahar were almost entirely excluded from literacy. Furthermore, urban environments provided a distinct advantage, as literate males were three times more common in cities than in rural areas, while literate females were six times more common. 

The gender gap remained one of the most prominent features of the census, with female literacy recorded at a minuscule 21 per 1,000. Although this represented a 71% proportional growth since 1911, the absolute numbers remained low due to entrenched social customs like purdah and early marriage, which frequently caused girls to be withdrawn from school prematurely. 

Officials also identified a "relapse into illiteracy," noting that many primary school students lost their reading and writing skills shortly after leaving school due to poor instruction and irregular attendance. This suggested that the state’s school enrolment figures often provided a misleadingly optimistic view of functional literacy. 

The 1918 influenza epidemic hit the illiterate rural poor harder than the wealthier literate classes, therefore its selective mortality may have artificially inflated literacy gains in some provinces. The 1921 report captured a colonial society at the very beginning of a mass education transition.

The 1921 Census of India’s approach to language was heavily shaped by the Linguistic Survey of India, marking a transition from pioneering discovery to the classification of data based on established scholarly schemes. Marten acknowledged significant practical difficulties in collecting accurate linguistic returns, noting that widespread bilingualism and a lack of technical knowledge among both respondents and enumerators often clouded the data. 

Because the public was generally unaware of scholastic distinctions between dialects, minor changes in census instructions could cause dramatic fluctuations in statistics; for instance, the recorded number of Dogri speakers was nearly halved when a specific instruction was omitted in 1921. Consequently, Marten cautioned that these figures should be viewed as general indicators of linguistic distribution rather than precise scientific measurements. 

Despite these administrative challenges, the census categorized approximately 316 million speakers into a handful of major language families. The Indo-Aryan family, comprising tongues of the northern and central plains like Hindi and Bengali, accounted for roughly three-quarters of the population with 232.5 million speakers. 

The Dravidian family followed as a massive second group with 64.1 million speakers, led by Telugu, which was the largest single Dravidian language.

The census also tracked smaller families along the subcontinent's fringes, including the Tibeto-Chinese (12.9 million), Austric (4.5 million), and Karen (1.1 million) families, which represented the diverse linguistic frontier of the empire. While the census documented a vast number of tribal and minor languages, many of these remained difficult to capture accurately due to the instability of local returns.

Religion was recorded from a simple instruction to enumerators: enter whatever religion each person returned, with tribal names recorded for those following none of the recognised faiths. Marten used this chapter to make a broader point about what such a return can and cannot mean. He observed, for most people, religion in the census sense was not a matter of doctrine but of communal identity; “that particular outlook and attitude towards the universe and his fellowmen which forms the tradition of his family and his clan.”

Of the population recorded, only about 850 individuals returned anything resembling agnostic, atheist or freethinker, not because personal doubt was rare, but because the census didn’t about personal beliefs.

Tribal Religions

One of the more self-critical commentaries in the report concerned what earlier censuses had labelled “Animism.” Marten renamed the category  as “Tribal Religions,” arguing that “Animism” implied a belief system the return never actually captured, and that the figures were of doubtful statistical value regardless. 

Their classification depended almost entirely on an enumerator's impression of whether a person of ambiguous identity should be recorded as Hindu or under their tribal name. Mr Sedgwick of Bombay put it bluntly: “our returns of Animists are absolutely worthless. They represent nothing and are entirely a matter of chance.” Marten retained the category chiefly to mark a population that definitely stood outside the circle of any of the main Indian religions. However he also conceded that its precise size was not reliable.

Hindu and Muslim population trends

The Muslim population grew by 3.1 percent over the decade against a slightly declining Hindu figure. This difference was attributed not to conversion, said to be by this period negligible, but to social custom. Islamic practice placed fewer restrictions on the remarriage of widows, and early marriage, while still common, was not so common as among Hindus, Marten noted. 

The Superintendent of Kashmir makes a related point, contrasting a Balti Muslim community practicing polygamy with Buddhist neighbours who shared wives among brothers to keep family estates intact. This illustrated how religious doctrine on marriage could shape a community's demographic and economic trajectory alike.