ISSN (Print) - 0012-9976 | ISSN (Online) - 2349-8846

A+| A| A-

Electoral Politics and the Manipulation of Statistics

The 2009 Lok Sabha election campaign has witnessed political parties making widespread use of constituency-wise data on economic and social indicators to attack each other and central/state governments. They have included statistical "evidence" in their manifestos and a number of large media outlets have cited these "data" as part of their efforts to educate the voter. Most unfortunate is that the mass of data on social and economic indicators provided by independent and private agencies to the media is of very doubtful value. These data do not appear to be based on any official/reliable sources, are inconsistent and have been created for years and for geographical units where no such data have ever been known to exist.


Electoral Politics and the Manipulation of Statistics


The 2009 Lok Sabha election campaign has witnessed political parties making widespread use of constituency-wise data on economic and social indicators to attack each other and central/ state governments. They have included statistical “evidence” in their manifestos and a number of large media outlets have cited these “data” as part of their efforts to educate the voter. Most unfortunate is that the mass of data on social and economic indicators provided by independent and private agencies to the media is of very doubtful value. These data do not appear to be based on any official/ reliable sources, are inconsistent and have been created for years and for geographical units where no such data have ever been known to exist.

Himanshu ( is at the School of Social Sciences, Jawaharlal Nehru University, New Delhi and also Visiting Fellow at Centre de Sciences Humaines, New Delhi.

lections provide an opportunity for citizens to make choices on the political representation that they want. While such choices should ideally be based on an independent evaluation of the performance of governments based on credible data, political parties and media also play a role in shaping opinion with the use of statistics on past achievements in the economic and social arena.

The 2009 Lok Sabha election has been no exception. Both the political parties as well as the media have used official statistics to evaluate the performance of central and state governments. The use of hard data to assess the performance of governments is a welcome development in electoral politics, particularly when the focus of an election is expected to be on development and governance. Unfortunately, some of the propaganda material used by political parties and the data presented by the media in the current campaign suggest a lack of honesty in using official statistics and in some cases even indicate a manufacture of evidence to suit political ideology.

For example, even before the election process got underway, reports were released in the media about the performance of the Left Front government in West B engal on the basis of a study sponsored by the main opposition party in the state, the T rinamool Congress (Debroy and Bhandari 2009). The authors used official data as well as their own estimates to argue that the Left Front government in the state had failed over the last 30 years.1 While the contents of the report were later challenged by the CPI(M),2 the purpose of the Debroy-Bhandari paper was served since the media carried the findings of the report in great detail and largely ignored the rebuttal.

Rise in Poverty

Around the same time, an article in the Mint reported the supposed findings of a study on poverty and human development based on district-level statistics.3 The study, undertaken by Buddhadeb Ghosh of Indian Statistical Institute (ISI), Kolkata, and sponsored by the Ministry of Statistics and Programme Implementation, covered all the districts of the country. However, the newspaper article gave specific attention to Murshidabad district of West Bengal.

In the same article in Mint, the author of the study was quoted as saying “By now, according to statistical estimates, the figure (poverty) has gone up by 20% (presumably since 2004-05), or by about fiveand-a-half crore (55 million) people”. Although the study is a work in progress and the final report has not been submitted, this single newspaper article was used in different ways without verification by both the Congress and the Bharatiya Janata Party (BJP) in their official publications.

The BJP used the “information” about the increase in poverty in its press briefings, in its election manifesto, and various leaders including the prime ministerial candidate L K Advani used this statement in many speeches.

Caught on the back foot, the Congress Party did try to check the veracity of this report, upon which the principal investigator of the study Buddhadeb Ghosh d enied having made any such statement.4 Even so, the same Congress which had e stablished that this media report on the all-India trend in poverty was not correct then used the same newspaper article to argue the failure of the government of West Bengal! Finance Minister Pranab Mukherjee released a white paper on the 30 years of Left Front rule which used the same “study” to argue that Murshidabad was the poorest district in the country.5 Remarkably, these claims by both parties were based on a single newspaper report, and neither saw it fit to undertake an independent verification of the data which are in the public domain.

National Sample Surveys

As is well known, all official estimates of poverty are based on consumption data collected by the National Sample Survey Organisation (NSSO) during the periodic national sample surveys. For example, on

Economic & Political Weekly

may 9, 2009 vol xliv no 19


the incidence of poverty after 2004-05, a recent paper published in this journal, u sing National Sample Survey (NSS) data relating to a “small” annual round for 2005-06, concluded that poverty did not increase between the two years and that in fact it declined slightly between 2004-05 and 2005-06 (Datta 2008). Since then, there has been another annual NSS round (2006-07) for which the basic results have been published. Estimates based on aggregate data collected in this round also show no increase in poverty. There are no other estimates of consumption expenditure that have been published by the NSSO. So the question then arises: what is this source of information for a study which claims that the number of poor has increased by 20% or 5.5 crore since 2004-05? It turns out that the ongoing study by B uddhadeb Ghosh is based on data only until 2004-05, and therefore it can make no claims regarding changes in poverty since then.

Similarly, the claim made by the white paper of the Congress Party that Murshidabad was the poorest district of the country is also belied by the available data and independent academic estimates. In a recent study of district-level poverty, also published in this journal (Choudhary and Gupta 2009), Murshidabad does not figure either in the worst 50 districts when ranked by the rural poverty headcount been at pains to clarify the correct position with respect to the all-India data on poverty and therefore clearly have access to the sources of the ongoing Ghosh study which also provide the statistics noted above. Even so, they have continued to use wrong numbers and false newspaper reports to buttress their criticism of the economic performance in states and districts ruled by the opposition parties.

Media Abuse

Such behaviour may be expected from p olitical parties who wish to vilify their opponents even if it means using false s tatistics. But what about the media and i ndependent research institutions? Two large English language media groups – Hindustan Times and Bennett Coleman – have put up information by electoral constituency on their web sites.7 The source of data for both is Indicus Analytics, an independent research organisation. Hindustan Times has also created a special web page in partnership with Google.8 This information is also available on the Indicus web site (

The Indicus data have also been used extensively by Indian Express for stories during the current election on constituencies. On the face of it, this can be a valuable Faridkot, Amethi, Chennai Central, Bangalore South, Ghatal, P onnani, Hajipur, Raebareli, Gandhinagar, Chhapra, Baramati, South Calcutta, Sivaganga, Hassan, Anantnag, Srikakulam and Bankura.

But how authentic are the economic data compiled by the independent private agency, Indicus Analytics, which have been used so extensively by so many m edia organisations?

What Do ‘Constituency’ Data Say?

This can first be illustrated with information on a few socio-economic indicators from the agency’s web site for a sample of high-profile districts. Table 1 provides some major indicators from the Indicus web site for Amethi (Rahul Gandhi), Raebareli (Sonia Gandhi), Jangipur (Pranab Mukherjee), Khammam (Renuka Choudhary), and Sivaganga (P Chidambaram) – all Congress. Then for Lucknow (Atal Behari Vajpayee) and Gandhinagar (L K Advani), both represented by the BJP, Ghatal (Gurudas Dasgupta) by the CPI, Bankura (Basudeb Acharia) by the CPI(M) and Baramati (Sharad Pawar) of the N ationalist Congress Party.

While Indicus provides data on a number of indicators, Table 1 gives information on total literacy, infant mortality

Table 1: ‘Statistics’ on Various Indicators in Some High-Profile Constituencies (2004 and 2008)

Sivaganga Amethi Raebareli Khammam Jangipur Indicator/Year 2004 2008 2004 2008 2004 2008 2004 2008 2004 2008

r atio and or among the 200 districts when Literacy (total) % 74 78 59 64 58 63 61 66 58

ranked by the urban poverty headcount BPL population % 24 22 47 44 53 58 28 30 53 53

ratio. A similar result also emerges from Fully immunised children % 84 79 20 16 20 16 60 57 41 47 the ISI work in progress. Infant mortality rate (deaths per 1,000 children under

Following a rebuttal by the CPI(M) of

one year of age) 36 32 88 83 86 79 42 41 65 62

this and other charges made by the

Per capita annual C ongress with respect to West Bengal, the expenditure (in Rs) 20,770 14,243 11,274 6,156 10,141 6,050 21,040 16,847 10,743 14,788

latter released another document6 in Annual household income (in Rs) 78,632 83,487 48,666 51,447 44,139 46,912 81,331 89,633 77,588 83,924

which the claim was made, once again on

Lucknow Gandhinagar Ghatal Bankura Baramati

the basis of the Ghosh study, that 14 of

Indicator/Year 2004 2008 2004 2008 2004 2008 2004 2008 2004 2008

West Bengal’s districts fall under poorest Literacy (total) % 71 75 80 82 76 78 59 63 83

100 districts of the country. Once again BPL population % 23 18 21 17 24 20 18

this observation is not supported either by official data or studies by academic r esearchers. The Choudhary-Gupta (2009) study finds only three districts of West Bengal among the poorest 100 in terms of rural poverty and no district of that state figures even in the 200 poorest districts in terms of urban poverty.

It is worth noting that several leaders and spokespersons of the Congress have

Fully immunised children % 35 34 51 45 67 79 71 84 38

Infant mortality rate 65 65 52 50 47 43 49 46 28

Per capita expenditure 18,820 11,285 25,680 34,254 17,619 16,468 11,382 13,091 32,884 36,530

Annual household income 1,02,884 1,13,381 1,67,163 1,88,515 82,411 89,279 76,349 82,939 1,86,257 1,99,805


data base that can enable common citizens rate, percentage of fully immunised chilto analyse the performance of their elected dren, below the poverty line (BPL) popularepresentatives. Outlook (4 May 2009) too tion, per capita expenditure and annual used this source to analyse the performance household income.9 The data are for two of various “high-profile” constituencies: years, 2004 and 2008.

may 9, 2009 vol xliv no 19


These statistics indicate a mixed p attern of progress in socio-economic i ndicators across constituencies. The only trend common to all of them is that literacy rates have risen uniformly. Yet, there are some odd patterns which must make us question the quality of the data that has been provided to and used uncritically by so many media organisations.

Per capita expenditure declined between 2004 and 2008 in six of the constituencies covered here while it increased in only four of them. At the same time household income has increased in all of them! The only possible explanation could be that people in constituencies where per capita expenditure has declined (Raebareli, Amethi, Sivaganga, Khammam, Finally, I added Bihar because it is often considered to be the “worst” state and there have been recent claims of revival after years of stagnation. Table 2 gives the same indicators as Table 1 except for the infant mortality rate which is surprisingly not available on the Indicus web site for states, although the same is available by constituencies.

State-level Data

Table 2 raises similar questions on the data and new ones as well. In child immunisation, rates have gone up between 2004 and 2008 in West Bengal, Assam and Bihar, while in the supposedly better administered states of Andhra Pradesh and G ujarat, they have actually declined. On

Table 2: ‘Statistics’ on Various Indicators –Sample States (2004 and 2008)

Andhra Pradesh West Bengal Assam Bihar Gujarat Indicator/Year 2004 2008 2004 2008 2004 2008 2004 2008 2004 2008

are rare to find in official statistics. They are typically available – if at all – with a long time lag or are not available for e very year.

Moreover, what we are being given by independent agencies like Indicus is constituency-wise information and since Lok Sabha constituency boundaries often criss-cross district boundaries, the data base for parliamentary seats will have to be based on an even more disaggregated information – block level data – that is then aggregated to provide data for individual constituencies. But, while there are limitations at the district level for various indicators there is certainly no database at the block level for most indicators on an a nnual basis.

The census data, which can yield district and block level information, is decennial and any extrapolation of census data

since the last one in 2001 is certainly no

Literacy (total) % 64 69 71 75 66 69 50 53 72 75 BPL population % 15 14 24 21 20 16 43 40 17 15 reflection of the current ground reality.

Fully immunised children % 48 41 62 71 30 37 30 40 46 42 And any extrapolation based on trends

Per capita annual
expenditure (in Rs) 16,680 19,087 18,512 19,472 13,336 12,722 10,433 5,444 21,009 27,143
Annual household
income (in Rs) 99,295 1,09,702 83,737 87,210 80,613 86,523 41,720 43,203 1,54,913 1,76,664


Ghatal and Lucknow) are saving more and consuming less despite the fact that many of them live in abject poverty. Or, just as unlikely, the average size of the household has changed dramatically between the two years, so much so that even as incomes have risen per capita expenditure/ incomes have fallen.

The other unusual aspect of this data is that immunisation rates, other than showing an increase in the three constituencies in West Bengal, have actually fallen in the rest! This too is unusual, for while India’s record in immunisation remains poor, no agency – official or independent – has pointed out that there has been an actual decline in the immunisation rate.

Given the problems at first sight with the data on some of the indicators it is worth looking at the same indicators at the state level. Among states, I have selected states with governments headed by the Congress (Andhra Pradesh), BJP (Gujarat) and Left Front (West Bengal). I have also included Assam, which is a Congress-ruled state and is also the “home state” of the present prime minister for the last 18 years.

economic indicators, all states show a decline in the BPL population percentage but surprisingly Assam and Bihar show a decline in per capita expenditure as well. Despite per capita expenditure declining by almost half, Bihar has done well to reduce the percentage of poor population by three percentage points.

The state-wise data raise their own questions. Going by this set of statistics, the performance of West Bengal is not a failure, at least compared to either the BJP-ruled Gujarat or the Congress-ruled states of Assam and Andhra Pradesh. However, the point is that the same set of data was also the basis of the claim made by Debroy and Bhandari (2009) of the failure of governance in West B engal, in a r eport published by the same institution that has provided the data on its web site. The question then is whether we can really trust these statistics.

If these statistics are actually true and reliable then they are an important addition to our knowledge base, and the firm has done a commendable service. This is especially so because district-level data b etween the 1991 and 2001 censuses is at best a reflection of the trend in the 1990s in districts and blocks.

All this suggests that there is a need to carefully examine the data made available by Indicus to various media houses, which the latter have used without assessment.

It does seem that there are several reasons why we should reject these statistics.

Obvious Problems

Some problems with the data are very

o bvious and are discernible when we just look at the trends. For example, unlike household income, per capita consumption expenditure does not fall drastically. Even if it falls in a year due to a severe c alamity, consumption expenditure is very unlikely to have fallen sharply in four years as some of the data in Table 1 indicates. As pointed out earlier, it also does defy the imagination that immunisation rates have fallen in all the constituencies listed in Table 1, other than those in West Bengal. Even as we accept that progress on this front has been slow, one can safely argue that there has not been a massive retrogression. There are also logical contradictions in the data presented by Indicus. How does one explain the apparent reduction (Table 2) in 2008 of per capita e xpenditure in Bihar to half of its level in

Economic & Political Weekly

may 9, 2009 vol xliv no 19


2004, when other data from the same source suggest that Bihar still manages to reduce the percentage of population which is poor?

A Sample Investigation

Fortunately, some of the data provided by Indicus can actually be cross-checked against official statistics. This is done as an illustration for three of the indicators.

(i) Per Capita Expenditure: The only known source of information is the consumption expenditure surveys conducted periodically by the NSSO. Of these, only on this have been presented in Choudhary and Gupta (2009).

It is apparent from a comparison of the Choudhary-Gupta estimates and that of Indicus that the latter has not only overestimated per capita expenditure but also throws up a different ranking of districts than the ones presented by Choudhary and Gupta using NSS data. For example,

Table 3: Per Capita Annual Consumption Expenditure

(in Rs, at current prices)

e xpenditure of as much as Rs 21,040 for Khammam.

After the 2004-05 NSS quinquennial round the latest annual NSS survey for which results are available is for 2006-07 but it had a very small sample so the data generated therein is not useful for making reliable district level estimates. And in any case unit level data from this round from which district level estimates can be calculated are not yet

2004-05 2006-07

available. The published NSS

Rural Urban Total Rural Urban Total

report for 2006-07 cannot be

Andhra Pradesh 7,247 13,097 8,744 8,724 16,332 10,605

used to calculate district level

Assam 6,921 13,555 7,534 8,652 16,428 9,345

per capita expenditure, but

Bihar 5,340 8,754 5,656 6,492 10,380 6,810

the quinquennial or large sample rounds Gujarat 7,738 14,470 10,060 9,564 17,064 11,858 the state level estimates in the

command some respectability for district West Bengal 6,908 13,908 8,621 7,560 16,452 9,304 report can be used to check the

level estimates, but even in those rounds there is no information at the block or village level which would make it possible to aggregate data into constituency-wise i nformation. The last NSS quinquennial survey that was carried out was in the 61st round (2004-05), and district level estimates of consumption expenditure based

NSS Reports on Consumer Expenditure (Report No 508 for 2004-05 and Report

validity of the Indicus data. As

No 527 for 2006-07).

can be seen from Table 3, the based on the NSS data, the per capita estimates from NSS published reports are e xpenditure for Khammam estimated by not only different from what is reported by Choudhary and Gupta is Rs 6,360 and Indicus for the selected states, the rankings Rs 9,516 in 2004-05 in the rural and urban too are different. For example, the NSS estiareas of the district. Compared to these, mates show that Andhra Pradesh has a Indicus reports an annual per capita higher consumption expenditure than West

SAGE half page

may 9, 2009 vol xliv no 19


B engal for both years. But the Indicus data says otherwise. Per c apita expenditure reported by Indicus is almost double that of the official NSS for most of the states. It is also clear from the NSS data that none of the selected states have suffered a decline in per capita c onsumption (all the states in fact showing a significant increase in per capita expenditure) unlike the Indicus data which shows per capita expenditure d eclining in Bihar and Assam.

  • (ii) Estimates of BPL Population: The same problem as for the per capita expenditure applies to the Indicus estimates of the BPL population, which are also usually based on the consumption expenditure survey of NSSO. There is one possibility and that is that the Indicus estimate is actually d erived from the BPL census conducted by each state. But even then it is impossible to have two estimates for 2004 and 2008, as Indicus has presented, since the only BPL census which is currently
  • o perational was that conducted in 2002. Based on that survey it is difficult to make any inference on the changes in the BPL p opulation in 2008. What is also interesting is that Indicus has apparently been able to measure the increase or decrease in BPL population at the block level without any surveys to assess these changes.
  • Perhaps it is time then to wind up the c ostly and controversial NSS consumption surveys for measuring consumption e xpenditure and poverty, since there seem to be other and more up to date information that data agencies like Indicus are able to tap.

    (iii) Immunisation Rates: The data on i mmunisation at the district level is now available from the third District Level Household Survey (DLHS-3),10 which is conducted by the Ministry of Health and Family Welfare. Along with data on health related indicators, DLHS also provides various basic aggregates at the district level through its District Level Household and Facility Survey.11 The last such survey was conducted in 2002-04. Unfortunately, the district level estimates presented on the Indicus web site are very different from those reported by DLHS-3 or by DLHS-2. For example, DLHS-3 shows an immunisation rate of 33.1%, an increase from 22.4% in DLHS-2 (2002-04) in Raebareli, as c ompared to a d ecline from 20% to 16% b etween 2004 and 2008 as shown on the Indicus web site for the district/constituency of that name. Of the results released so far by DLHS-3, not a single district shows a d ecline in immunisation coverage, but the Indicus data show that all constituencies have seen a decline. Even the absolute levels of immunisation as reported by Indicus are at variance with the DLHS data. Similar discrepancies are observed in the data reported for many other constituencies.

    There are two other sources that Indicus could have used for data on immunisation rates: the Sample Registration System (SRS) and the National Family Health S urvey (NFHS).

    The NFHS is available for 2005-06 but data are available only at the state level. That leaves the SRS which is available d istrict-wise from the census office. But here again, the latest data available is for 2006 and not yet for all districts, and there are no block level data that could be used for constituency mapping. In any case the Indicus data does not match the SRS data either.

    It is worth noting that Indicus does offer a disclaimer on its web site, “The contents are for information purposes only and our estimates using latest available data from highly credible sources. Although every care has been taken in their compilation, neither Indicus Analytics nor its employees accept any responsibility whatsoever for any consequences of their use” ( emphasis added).

    In a description of the data set, it is mentioned on the Indicus web site that the data are based on government sources, a lthough nothing is divulged on the actual sources. However such disclaimers do not abdicate responsibility for authenticity, especially when there is so much variation between the official and the Indicus data and when it is not clear where the data comes from for recent years.

    Fictitious Data?

    On the whole, a careful check of the data on the Indicus web site suggests that most of the data are either fictitious or make a questionable extrapolation of old data. In either case, the data lack credibility. Data projected on the basis of the censuses of 1991 and 2001 cannot be a good indicator for 2004-08. For any serious researcher, such data as put out by Indicus cannot be an authentic source of information. N evertheless, such information is now b eing regularly used by media professionals to substantiate their statements about the failures and successes of political parties and governments.

    It is necessary to provide information to the electorate so that they can make informed choices. But such a service should not be at the cost of compromising the trust that the electorate reposes in the m edia and serious researchers. The least that organisations like Indicus could have done was to report its definitions and data sources clearly, which is the practice with all rigorous and reliable exercises in crosscountry comparisons such as the World Development Indicators, published by the World Bank.


    1 “Bengal Needs Better Governance, Says Trinamool-funded Report”, Times of India, 27 February 2009.

    2 People’s Democracy, 16 March 2009. 3 “Once-prosperous Silk Centre in a Sorry State”, Mint, 15 February 2009. 4 “No NSS Data on Poverty, Jairam Rebuts Sudheen’s Claim”, 2009/apr/03/jairam-ramesh-rebuts-kulkarnisclaim.htm 5 “30 Years of Left Front Rule in West Bengal: A Development Report Card”, available at www.aicc. 6 “The Left Front’s Un-development Report Card Part II: The Truth behind the Left Front’s Fallacious Counterclaims”, available at 7 and

    tionspecial.cms. 8 9 Although it is not clear as to what “annual house

    hold income” represents, it appears that the data is actually of the district domestic product per household. It is also not clear why Indicus has chosen to present per household and not per c apita data, which is the standard practice when providing information on income/expenditure for such indicators.

    10 District Level Household Survey by the Ministry of Health and Family Welfare is one of the largest demographic and health surveys in India covering all districts of the country.

    11 Although there are concerns about the reliability of DLHS data at the district level, this is the only source of information on many indicators for 2007-08 at the district level with a sample size of seven lakh households.


    Choudhary, Siladitya and Nivedita Gupta (2009): “Levels of Living and Poverty Patterns: A District Level Analysis”, Economic & Political Weekly, 28 February.

    Datta, K L (2008), “An Estimate of Poverty Reduction between 2004-05 and 2005-06”, Economic & P olitical Weekly, 22 November.

    Debroy, Bibek and Laveesh Bhandari (2009): “Transforming West Bengal: Changing the Agenda for an Agenda for Change”, available at

    Economic & Political Weekly

    may 9, 2009 vol xliv no 19

    Dear reader,

    To continue reading, become a subscriber.

    Explore our attracive subscription offers.

    Click here


    (-) Hide

    EPW looks forward to your comments. Please note that comments are moderated as per our comments policy. They may take some time to appear. A comment, if suitable, may be selected for publication in the Letters pages of EPW.

    Back to Top