The dumbing down of data
Census data on religion collected in February-March 2011 was belatedly released on August 25, 2015. This analysis of how the media covered the release and presented the data will first give an overview of the coverage in five English dailies and then look at the shortcomings.
Our governments spend enormous amounts – Rs 2,200 crores on the 2011 Census – to collect data that they are reluctant to release, let alone use. A comparison of five English dailies published from Bangalore – The Hindu, Deccan Herald (DH), and The Times of India (TOI) –and New Delhi – Hindustan Times (HT) and The Indian Express (TIE) –suggests that the religion data did not receive sufficient coverage in the print media. This is surprising as census data is released once in a decade and we have a long history of politically-motivated use of community-level statistics. The Census seems to have been crowded out by Indrani Bora and Hardik Patel. We will first compare the coverage – front page, other pages/supplements, opinion pieces, and editorials – of the latest census data in different newspapers before discussing their shared shortcomings.
First day-first page
On August 26, three newspapers carried the news about the census data as the lead story on their front pages, while DH and TIE placed this in the lower half of the first page.The headlines of three of them highlighted the demographic “imbalance” – TOI (Muslim share of population up 0.8%, Hindus’ down 0.7%), HT (Hindu proportion of India's population less than 80%), and DH (Muslim population grew faster: Census) – whereas The Hindu (Muslim population growth slows) highlighted the declining rate of growth rate of Muslim population. The TIE headline (Hindus dip to below 80% population; Muslim share up, slows down) highlighted both the “imbalance” and declining growth rate of Muslims. HT noted that the growth rate of the Muslim population has slowed down, the share of Muslims in Gujarat’s population has dropped, and the BJP released the data, which should have been released by the UPA-II, just ahead of the Bihar polls.
HT also suggested that the government did not release cross-tabulated data correlating religion and socio-economic indicators to minimize public debate/controversy. DH drew attention to the fact that government released religion rather than caste data ahead of the Bihar Assembly elections. The Hindu drew attention to the belated release in a communally surcharged political environment. It noted that the Hindu-Muslim growth rate gap has narrowed because the Muslim population’s “growth rate has slowed more sharply”and added the relatively higher growth rate of Muslims can be explained by “higher Muslim fertility, higher child mortality among Hindus and a greater life expectancy among Muslims.” TIE noted the difficulty in reading the data on communal lines and linked the release to elections in Bihar, Assam, and West Bengal, “three states with a significant Muslim population.” TOI highlighted the large jump in “the number of people who did not state their religion.”
First day-other pages
The first day’s coverage extended beyond the main article on the first page in all the newspapers, except TIE whose coverage picked up momentum on August 27. HT (HT Spotlight/Decoding Census) and TOI (Times Special/Divine Numbers) carried full page supplements including extensive graphics and tables. TOI’s supplement carried five articles that highlighted the concentration of Christians in south India and religious minorities in urban areas, the increase in the number of Muslim majority districts in Assam, and explored the relationship between religion, region, and sex ratio. It used saffron and green in the graphics to depict Hindus and Muslims. HT’s supplement speculated that the marginal increase in the Muslim population share could affect the political discourse centered around caste, highlighted the debate about the Bangladeshi influx in Assam and West Bengal, and clumsily blamed economic backwardness for the high population growth rate of Muslims in Kerala.
The Times of India (Hindu population rose marginally in Karnataka, August 26, Page 16) also published a news item in addition to the supplement. DH carried an article on Page 8 (Christians in India have more women than men) comparing the distribution of Christians and Muslims. The Hindu carried an article (Muslim sex ratio improves further, August 26, Page 12) highlighting the better sex ratio among Muslims. Other news reports published in The Hindu on Page 1 (Bihar elections among factors in data release) and Page 12 (Lok Sabha elections deterred UPA) suggested that while the UPA “did not want to touch the data in view of the general elections in May 2014,” the BJP that was expected to release the data “once it assumed office” failed to release the data due to “lack of political will.”
Other days
The Hindu alone covered the issue on its first page on August 27 (Religion data released with little warning, no context) and claimed that the release “marked by secrecy” was “a departure from the norm.” It added that “The press release gave no historical context, and the data on the census website was mere numbers, with no explanation” and that this was “intentional” as the Registrar General has adopted the “policy of only releasing data online henceforth.” But it added that in 2004 the Registrar General released the data with great fanfare only to find himself in the middle of controversy.
DH carried stories on August 27 (Hindu population grew by over 13 crore in 10 years, Page 8), which suggested that the statistics did not support the fear of “unchecked” population growth of minorities, and August 29 (Think tank debunks fear over rise of Muslims, Page 9), which highlighted the decline in fertility across religions and added that the “Fertility among Hindus of UP is higher than that of Hindus of Tamil Nadu and the same holds true for Muslims.”
A number of news articles in TIE (August 27, Page 10: Aditya Nath, Sakshi revive demand for Uniform Civil Code and ‘Govt. using census data to polarize Bihar before polls’; August 29: Rising Muslim population a worry, says Togadia, Page 13; August 31: Sounding a note of caution, Page 5; ‘Hindu women should not marry outside community,’ Page 7; Assam Muslim growth is higher in districts away from border, Page 8) covered the communal propaganda related to religion data.
In addition to the news reports discussed above, the newspapers carried very few other reports related to the latest census data on religion or on the controversies triggered by that data – DH (Gurus ask Sikhs to have more than three kids, September 04, Page 12), HT (Togadia, Sakshi target Muslim population spike, September 04, Page 11), and The Hindu (Region and religion both matter for better population indicators, August 28, Page 14). "HT reported that Togadia suggested that access to government welfare schemes should be made conditional upon family size and violation of two-child norm should be criminalised ...", whereas the article in The Hindu pointed out that “The population growth rate for Muslims in Kerala . . . [is] half that of Hindus in States like Bihar.”
Opinion pieces
Only The Hindu did not publish any opinion piece on this issue. TOI carried an article (Census, Maps, Human Presence and Planetary Imperatives, August 28), which peripherally engaged with the Census and suggested that by focusing upon ever smaller fractions of mankind we might be looking “more and more on details that might be pretty insignificant when seen from a planetary, galactic or universal perspective.” HT’s opinion piece (Don't turn census data into half-truths, August 31) examined Islamic perspectives on birth control and also noted that “the Muslim growth rate in educated Kerala is lower than all the communities of northern states, there by proving the thesis that fertility rates have nothing to do with religion. Education, financial status, health care and regional factors are far more important than religion” (emphasis added).
Hindustan Times also published another opinion piece (A social fabric that will hold, September 03) that briefly referred to the Census before discussing Yakub Memon’s hanging. DH belatedly published an opinion piece (Spreading misinformation, September 05) that argued that forces transcending religious boundaries are driving down fertility across the country and dismissed the unscientific fear that Hindus will be swamped by Muslims.
The Indian Express published five opinion pieces.The first opinion piece (Population growth slowing for all; on sex ratio, Muslims better than Hindus, August 27) debunked communal interpretations of the data. Another piece (Myth of Muslim growth, September 02) revisited the controversy around the release of the 2001 data and its relation to the Sachar Committee and then explored a variety of factors that affect fertility. Three other opinion pieces (Census, Christians, conversions, August 29; Bend it like Bhalla, September 01; Numbers do not lie – people do, September 05) that debated the role of conversion in the growth of the Christian population were notable for the selective use of data. All three pieces overlook the inter-relationship between tribe/caste and Christianity in India and also fail to note that the data on tribes is the most unreliable part of our censuses.
Editorials
The Hindu, TOI, and HT addressed the issue in an editorial on August 27, whereas DH waited until August 31 to take an editorial stand and TIE skipped this issue. TOI (Region, Not Religion) used the editorial to shift to the middle ground. It highlighted the decline in fertility cutting across communities and the complex relationship among religion, region, socio-economic development, governance deficit, and fertility. HT’s editorial (Time to figure it all out) summarized the first day’s stories and emphasised the Bihar angle. It concluded that other minorities “run the risk of becoming forgotten minorities as Muslim numbers grow.” DH (Don’t politicize population figures) advised against reading the data along community lines and used the editorial to reaffirm the more nuanced position adopted in the coverage after the first day.
The Hindu’s editorial (Questions of timing and context) reinforced concerns about the context and the manner of the release and pointed out that “the government did not step forward to take institutional responsibility” for releasing communally sensitive data possibly because this suited the ruling party. The coverage in The Hindu was capped by an article by the Readers’ Editor (Making sense of metadata, August 29) who drew attention to the newspaper’s “in-house expertise in data journalism” and suggested that “In The Hindu’s editorial judgment, the Census was primarily a demographic document and it decided to look at the findings through a demographic lens.” The article concluded by claiming that by placing data in its context The Hindu “blunted the power of the polarising agents,” which reminds of TOI’s claim: “You read it first in TOI . . . on Jan 22 – the Muslim population grows 24%, but slower than the last decade.”
A few general points need to be noted before we critically examine the coverage. The Hindu relied upon a number of sources including professional demographers and government officials for its news reports, while TIE relied upon opinion pieces by experts. The Hindu did not publish any opinion piece, whereas TIE did not publish any editorial on this issue. TOI’s captions stand out: “Divine Numbers,” “Census: A faithful picture of India,” “Religious state of the states,” and “Bengal beats India in Muslim growth rate.” Interestingly, none of the newspapers carried any cartoon on politics over religion data.
SHORTCOMINGS OF THE ANALYSIS
The context of the delay
Almost all newspapers suggested that the data were belatedly released, without presenting any timeline or indicating what constitutes a delay. They claimed that the UPA-II delayed the release due to elections and then referred to religion data leaked in January 2015 to suggest that the data was ready months ago. The Hindu (August 26) added that “The 2001 Census data on religion was released in 2004” and, therefore,“the 2011 round results were expected in 2014.”
Three points overlooked by the media would have strengthened the claim about the delay. First, the increasing use of information technology should have reduced rather than increased the delay. The religion tables from the March 1971 census were released in October 1972. (But, in the aftermath of Pakistan’s linguistic partition, the government released the language data in 1976 after reclassifying the smaller languages.) Second, in 2001 the Registrar General adopted new technology and had to additionally cater to the requirements of the Delimitation Commission. Unencumbered by such factors, the Registrar General could have released the data much earlier this time.
Third, the explanatory note to question eight of the 2011 Census Household Schedule clearly links the identification of scheduled castes to their religious affiliation that is canvassed in question seven. This means that the caste and religion data have to be sorted simultaneously. Information about caste (and even tribe) from the 2011 Census was available as early as April 30, 2013, i.e., a year before parliamentary elections, when the primary census abstract was purportedly released a year ahead of schedule. Since only stripped down Excel tables on major religions were released in August 2015, the government cannot justify the delay by arguing that it needed additional time after April 2013 or January 2015 to prepare a detailed report on religion or sort data about non-major religions.
The Bihar election factor
Now let us look at the political context. Here again we find our newspapers lacking. Three points bear noting. First, none of the newspapers examined the Bihar data to identify the channels through which the data could affect electoral outcomes. The drop in Bihar’s Hindu population share is not more than the drop in the national share of the Hindu population and Bihar’s Hindu population share continues to be well above the “psychological threshold” of 80 per cent. Whether the effect of data will be mediated through the general scaremongering of the Right wing or any Bihar specific demographic statistics can be ascertained only through an analysis of district level population changes.
Another aspect of the Bihar hypothesis needs attention. If the government had published the data after the Bihar elections it would have been accused of causing the additional delay on account of polls. And, at the same time, it would have attracted criticism for releasing the data ahead of elections in Assam, Bengal, and Kerala, states with above national average Muslim population shares. (Once there is a delay in releasing data it is impossible to find a politically non-controversial opportunity for release.)
Second, the delay has to be viewed within the context of the growing political interference in government statistics over the years, irrespective of the ruling party. We need to go back to the report of the National Statistical Commission, among others, and begin the hard work of institutional reform, strengthening the autonomy of both the agencies that collect data and those that contribute to transparency within the government. The issue of reform was completely ignored by the media.
Third, the newspapers did not examine the BJP’s statistical U-turn. The BJP had attacked the “anti-national” UPA-II for “suppressing census figures” because it was “ashamed to admit its failure to take the Muslims out of deep poverty.” The BJP’s manifesto hinted at the need to overhaul the government’s statistical machinery and presented real time and big data as means to solve problems as diverse as agricultural distress and intelligence gathering. Ironically, the party that promised real time data did not know how to release real data collected half a decade ago.
Demography
Our newspapers seem to have misunderstood the nature of the data and the kinds of claims that can be made using the data. For instance, the Reader’s Editor of The Hindu noted that “The journalistic process gets more complicated . . . when reporting metadata like the Indian Census figures” (emphasis added). He added that only “A systematic disaggregation enables us to understand the full implications of the individual strands — social, economic and regional among many others — that intertwine to form the complex metadata.” But metadata refers to data about data, information about information. While the Census also publishes metadata, the data released on August 25 were not metadata.
A few other points need to be highlighted. First, two newspapers highlighted the greater decline in Muslim population growth rate in absolute terms from the 2001 level. The rest, which initially highlighted the higher growth rate of the Muslim population, converged to the above point later. But one could as well claim that the growth rate of Hindus dropped more than Muslims relative to the level in 2001.Both statistics should have been shared with the readers to convey the message that the multiplicity of statistics that are not always in agreement with each other allows political parties to cherry-pick their favourite statistic.
Second, none of the newspapers doubted the reliability of the data. For instance, the HT supplement showed that the Christian population contracted in Nagaland between 2001 and 2011. The contraction is an artefact of massive over-reporting in 2001, when the overall population was inflated by as much as 30 per cent. The state government rejected the 2001 Census and vowed to conduct a clean census in 2011. This time ghost entries did not find their way into the census records leading to a drop in the overall population of this predominantly Christian state. What appeared as a contraction to HT was, in fact, a correction of the massive inflation in earlier censuses and is an example of how more reliable data can be collected by taking civil society into confidence.
Likewise the growth rates of Muslims in Jammu and Kashmir (26.1%) and Christians of Meghalaya (172.22%) need careful analysis. In the case of Meghalaya, HT published a grossly incorrect estimate of the growth rate of Christians as it used an incorrect figure for the 2001 population of Christians of Meghalaya, 812,961 instead of 1,628,986. Using the latter figure, which is reported in The First Report on Religion Data (published on Sep 06, 2004) based on the 2001 Census, the growth rate between 2001 and 2011 is 35.85 per cent.
The case of Jammu and Kashmir is a little more complicated. Guilmoto and Irudaya Rajan (EPW, June 8, 2013) have argued that the sudden drop in Jammu and Kashmir's child sex ratio can be explained by "The fear of a census conspiracy to inflate Jammu’s population could have induced people in the rest of the state, especially in the Kashmir Valley, to overstate their own household population by adding non-existent [male] children. "Their assessment is not implausible and hence a proper estimate of the state's community level growth rates needs to wait until a careful analysis of the child sex ratio is carried out.
On August 28, The Hindu (Region and religion both matter for better population indicators, Page 14) published a graphic that among other things showed that the tribal population of UP grew by 950.61 per cent between 2001 and 2011. This is incorrect. This growth reflects the reclassification as ST of about 10 communities (comprising of 18 community names), which were deliberately mis-classified as non-ST in the past. An estimate of ST population growth for 2001-11 should include the population of these "new" ST communities in the base year (2001) as well.
Third, north-south comparisons were made to judge the contribution of religion to fertility. A decade ago, under similar circumstances, Mari Bhat and Francis Zavier noted that “It is occasionally asserted that the fertility of Muslims in a state such as Kerala is lower than the fertility of Hindus in a state like Uttar Pradesh and, hence it cannot be claimed that Muslims have higher fertility than Hindus. But such an argument is absurd because factors other than religion also have an influence on fertility. Data from the NFHS also show that the fertility of illiterate women in Kerala is lower than the fertility of women who have completed high school in Uttar Pradesh. Could this evidence be used to refute the effect of female education on fertility?” (EPW, Jan 29, 2005)In other words, while there is a need to contest the right wing’s propaganda, the counter-narrative has to be built on a careful demographic analysis a la Bhat and Zavier and others.
Fourth, The Hindu’s editorial claimed that “All collections of data are value-neutral, but the manner of their release and potential for use and misuse are not.” Even a cursory familiarity with the Census would suggest the contrary. Every category used in the Census, the sequence in which questions are asked, and the format in which the data are printed are deeply informed by political and socio-economic considerations.
In Nagaland, the most widely spoken language (Nagamese) is not reported in the Census because it is tainted by Assamese. In the Household Schedule, the question on religion precedes the question on caste because it is presumed (with constitutional sanction, of course) that scheduled castes are confined to Hindu, Sikh, and Buddhist communities.
Before 1971, the Census used to publish detailed data on languages. Ganesh Devy has argued that the Census restricted the information on language in its publications "following the Bangladesh war when East Pakistan cited language as a reason to break away from West Pakistan" and "languages spoken by less than 10,000 people have been lumped as “others”."
Lastly, the Census prioritises "major" religions by first publishing data on six religions and then releasing data on "minor" religions," even though some of the "minor" tribal religions have populations comparable to the major religions. These examples provide a mere glimpse of how non-demographic considerations structure the whole exercise of the Census right from the stage of deciding what to ask and to when and how to publish the results.
Fifth, there was hardly any discussion about the non-release of language data, the decline in the Buddhist and Sikh population, the unusually high and low growth rates of Jains and Christians in certain states, the absence of data about the smaller religions, and the alleged role of Bangladeshi immigration in the population growth in Assam and West Bengal. Sixth, newspapers focused entirely on how census data could vitiate political processes, without examining the possible errors in the data or the impact of the deepening communal crisis since the mid-1980s on the quality of religion statistics.
To conclude, the decennial Census data on religion received superficial coverage in the print media. This is indicative of both the lack of serious interest in the matter beyond facile hypotheses about its purported electoral implications, as well as the underdeveloped state of data journalism. Given the growing use of statistics in almost every walk of life, our newspapers/channels ought to look beyond election data analysis and develop a serious capacity to engage with data.
(Vikas Kumar teaches economics at Azim Premji University, Bangalore.)