1.No excuses please, India awaits a full caste headcount
The central government’s logistical or legal justifications to not disclose caste census data do not stand scrutiny
The Narendra Modi government has finally elaborated on its reasons for not disclosing the caste data collated in the Socio-Economic and Caste Census (SECC) 2011. The Government’s affidavit filed in the Supreme Court of India last month by the Union Ministry of Social Justice and Empowerment — in response to a writ petition by the Maharashtra government — has also tried to explain away the exclusion of full caste enumeration in the forthcoming Census exercise, which is expected to resume in 2022 after the COVID-19 pandemic-induced interruption. The official arguments focus on the impracticability of full caste enumeration, suggesting that operational difficulties are simply too overwhelming.
Overstating ‘mistakes’, ‘flaws’
As detailed in the affidavit, while the total number of castes counted in the 1931 Census was 4,147, the SECC of 2011 returned over 46 lakh caste names. Such a humongous number of castes were returned partly because Indian people use the names of their caste, sub-caste, clan, gotra and surnames interchangeably. Moreover, the enumerators also got confused over the spellings and classification of the castes. The question is, if the 46 lakh caste names that were returned in the SECC 2011 were the results of “mistakes by the enumerators” or “inherent flaws in the manner of conducting census” as alleged in the official affidavit, why could not those mistakes and flaws be rectified in a decade?
The Union Cabinet had appointed an Expert Committee headed by the then NITI Aayog Vice-Chairman Arvind Panagariya in July 2015, charging it with the classification of caste names returned in SECC 2011. The Government’s affidavit admits that no other member was appointed to the committee. Therefore, neither did the committee ever meet nor did it fulfil its mandate in six years. Who is responsible for this negligence?
The extent of errors in caste enumeration is also being exaggerated. The affidavit cites the example of Maharashtra to show that while the aggregate number of castes enlisted in the central lists of Scheduled Castes (SC) and Scheduled Tribes (ST) and the State list of Other Backward Classes (OBC) is 494, the caste names returned by the SECC 2011 for the State stood at 4.28 lakh. But the affidavit also states that “99% of the castes enumerated had [a] population of less than 100 persons”. Out of the total population of 10.3 crore in the State, 8.8 crore people could be classified under 2,440 castes, each having a population of over 1,000. Therefore, the proliferation of caste names and the consequent difficulty in their classification have arisen not because of the majority, but a tiny proportion of the total population.
This is further corroborated by an action taken report on the recommendations of the Standing Committee on Rural Development on “BPL Survey (currently Socio Economic & Caste Census, 2011)”. In this report dated August 31, 2016, the Union government is quoted as categorically stating that the data errors on caste and religion in SECC 2011 accounted for 1.34 crore out of 118.64 crore people, i.e., only around 1% of the total enumerated population. How can the same government now characterise the entire findings of the caste census as “fraught with mistakes and inaccuracies”?
Complex, yet feasible
Population census in a vast and uniquely diverse country such as India cannot but be a complex exercise. Over the decades, the census machinery has moved on a learning curve, credibly enumerating complicated categories such as language and religion, which also display considerable diversity. For instance, as per Paper 1 of 2018 on the Language Census of 2011, the number of initial raw returns of mother tongues had totalled 19,569 for the entire country. Following scrutiny, editing and linguistic grouping, these raw returns were first rationalised into 1,369 mother tongues and subsequently classified on the basis of at least 10,000 or more speakers for 22 scheduled and 99 non-scheduled languages, i.e., 121 languages at the all-India level.
While caste appears to be an even more complex category than language in the Indian context, technologies to enumerate and analyse complex big data have become easily accessible today. Yet, the affidavit cites the absence of an all-India Registry of Castes to rule out the conduct of full caste enumeration in the forthcoming census. Why could such a registry of castes and tribes not be created till date by the Union and State governments, acting together, by combining the central lists of SCs and STs and the State lists of OBCs?
The aggregate number of castes and tribes included in those lists would currently be around 5,000 at the all-India level. For any individual State, the maximum number of castes cannot be above 500. Rationalisation and classification of the numerous raw caste returns into a maximum of 500 castes at the State level or around 5,000 castes at the all-India level, is eminently feasible. Training manuals for the enumerators can also be drawn up on the basis of a single, consolidated caste list for each State.
This could have been attained by the expert committee appointed by the Union cabinet by now, but for its innate dysfunctionality. Rather than rectifying its administrative failings, the Union government is now citing it as evidence to construct a theorem of logistical impossibility.
Within the framework
The Government’s affidavit also cites the absence of categorical constitutional or statutory requirements to count castes other than SCs and STs in the Census. However, Articles 15(4) and 15(5) of the Indian Constitution have explicitly recognised “socially and educationally backward classes of citizens” as a category distinct from SCs and STs and enabled the State to make special provisions for their advancement. Counting the population of these Backward Classes would therefore be very much within the constitutional framework.
Yet, the official affidavit alleges that full caste enumeration may compromise the basic integrity of the Census exercise, distorting the fundamental population count itself. If enumeration of individual castes under the “SC”, “ST” and “Other” categories in all censuses since 1951 have not led to such perverse outcomes, why should the additional enumeration under another “OBC” category cause such a catastrophe? Such deliberate scaremongering has no basis in the laws or the Constitution.
The logistical or legal justifications of the Union government to not disclose caste census data and refuse to conduct a full caste enumeration in the forthcoming Census do not stand scrutiny. Rather, it creates ground for suspicion whether the establishment has vested interests in concealing the real numbers and proportions of various castes in the Indian population. Such subterfuge would not be acceptable to a wide spectrum of social movements and political parties, who are demanding full caste enumeration.
Enumerating, describing, and understanding the population of society and what people have access to, and what they are excluded from is important not only for social scientists but also for policy practitioners and the government.
In this regard, the Census of India, one of the largest exercises of its kind, enumerates and collects demographic and socio-economic information on the Indian population.
However, the critiques of the exercise of the census consider it as both a data collection effort and a technique of governance, but not quite useful enough for a detailed and comprehensive understanding of a complex society.
In this context, the Socio-Economic and Caste Census (SECC) was conducted in 2011, but it has its own issues.
Census, SECC & Difference
- The origin of the Census in India goes back to the colonial exercise of 1881.
- Census has evolved and been used by the government, policymakers, academics, and others to capture the Indian population, access resources, map social change, delimitation exercise, etc.
- However, as early as the 1940s, W.W.M. Yeatts, Census Commissioner for India for the 1941 Census, had pointed out that “the census is a large, immensely powerful, but blunt instrument unsuited for specialized inquiry.”
- SECC was conducted for the first time since 1931.
- SECC is meant to canvass every Indian family, both in rural and urban India, and ask about their:
- Economic status, so as to allow Central and State authorities to come up with a range of indicators of deprivation, permutations, and combinations of which could be used by each authority to define a poor or deprived person.
- It is also meant to ask every person their specific caste name to allow the government to re-evaluate which caste groups were economically worst off and which were better off.
- SECC has the potential to allow for a mapping of inequalities at a broader level.
Difference Between Census & SECC
- The Census provides a portrait of the Indian population, while the SECC is a tool to identify beneficiaries of state support.
- Since the Census falls under the Census Act of 1948, all data are considered confidential, whereas according to the SECC website, “all the personal information given in the SECC is open for use by Government departments to grant and/or restrict benefits to households.”
Associated Concerns With SECC
- Repercussions of a Caste Census: Caste has an emotive element and thus there exist the political and social repercussions of a caste census.
- There have been concerns that counting caste may help solidify or harden identities.
- Due to these repercussions, nearly a decade after the SECC, a sizeable amount of its data remains unreleased or released only in parts.
- Caste Is Context-specific: Caste has never been a proxy for class or deprivation in India; it constitutes a distinct kind of embedded discrimination that often transcends class. For example:
- People with Dalit last names are less likely to be called for job interviews even when their qualifications are better than that of an upper-caste candidate.
- They are also less likely to be accepted as tenants by landlords. Thus difficult to measure.
- Marriage to a well- educated, well-off Dalit man still sparks violent reprisals among the families of upper-caste women every day across the country.
2.Getting to the heart of causality
The Economics Nobel Laureates have successfully upturned received wisdom in mainstream economics
Does the entry of immigrants reduce employment and lower the wages of native workers? Does the introduction of minimum wage, designed to protect workers, end up harming them by reducing employment? Does compulsory schooling affect schooling and earnings? If people got a basic income, would they stop working for a living? Graduates of private universities earn more than graduates of public universities in the U.S. Does this mean that attending private universities bestows a wage premium?
Drawing causal inference
These are among the many significant and deep questions that the three Economics Nobel Laureates for 2021 — David Card, Joshua Angrist and Guido Imbens — have investigated. Answering such questions involves establishing causality accurately. One way to draw causal inference is through experiments or randomised controlled trials (RCTs), the predominance of which in the field of empirical economics was recognised by the Nobel Committee in 2019. However, several big-picture and urgent questions cannot be evaluated through RCTs because of ethical, logistical or financial reasons.
Outside of experiments, researchers have to rely on real-world data which is messy. Drawing the correct causal inference entails comparisons between groups (those who stayed longer in school compared to those who did not, or States where the minimum wage increased compared to States where it did not increase, and so forth). But since individuals or States differ along many dimensions, comparison needs to be done carefully to avoid comparing apples with oranges. Additionally, adjustments need to be made for self-selection and omitted variables that might confound causal inference. The 2021 Nobel Laureates have been justly recognised for their pioneering contribution to the methodologies to uncover causality using real-world observational data. In doing so, their path-breaking studies have successfully questioned established orthodoxy and upturned received wisdom in mainstream economics.
Professor Card and Alan Krueger’s most influential 1992 study estimated the effect of minimum wage increases. The two economists utilised a ‘natural experiment’ (in which individuals are randomly exposed to a change caused by nature, institutions, or policy changes): in this case, a policy change in New Jersey that raised minimum wages for its low-skilled workers. Instead of comparing change in employment in New Jersey before and after the wage increase, as that could be affected by several other factors, they compared a double difference (‘difference-in-differences’): employment in New Jersey before and after the policy change compared to neighbouring Pennsylvania, where wages did not change, over the same period. Contrary to established wisdom, they found that an increase in minimum wages did not lead to a reduction in employment. This study has been replicated since by other researchers across several rounds of minimum wage increase and each time, the result has been the same, viz., no adverse impact on employment.
Why is the textbook prediction not borne out by data? There are many reasons: one is that the mythical perfectly competitive labour market, where firms are price-takers, i.e. they have no autonomy in wage setting, does not exist in reality. It turns out that monopsonistic firms (very large employers with market power) can set wages lower than the competitive wage and earn a surplus. Therefore, when the government imposes a minimum wage, the number of workers employed does not necessarily decrease but a part of the surplus now gets transferred to the workers.
A source of great anxiety in the contemporary world is the apprehension that entry of immigrants will adversely affect employment and wages of non-immigrant residents. Prof. Card’s analysis of another natural experiment — the Mariel boat lift that brought 1,25,000 Cubans to the U.S. in 1980, half of whom settled in Miami — showed this anxiety to be invalid. As a result of the boat lift, the Miami workforce increased by 7% but this had no adverse impact on the wages or employment of the non-Cuban native workforce.
Both Prof. Imbens and Prof. Angrist have made innovative methodological contributions to causal inference that have enabled explorations of big-picture questions. For example, in the U.S., private university graduates earn 14% higher wages than public university graduates. Does that mean private universities cause wages to go up? Prof. Angrist’s research corrected for ‘selection bias’, i.e. adjusted for the fact that SAT scores and family incomes are higher for private university entrants. Comparing like with like, the study finds that attending private universities does not confer a wage premium.
These causal techniques are based on a comparison of observed outcomes with counterfactuals: the ‘what if’ scenarios or ‘potential outcomes’ that are not observed. Such comparisons are logically compelling, and these methods have been replicated by hundreds of researchers across a variety of contexts, validating their effectiveness in disentangling causal effects from messy observational data.
Making metrics fun, relevant
The Nobel Laureates have also contributed to transforming pedagogy. Learning econometric techniques through the body of their work has made econometrics less abstract, more relatable and interesting. Prof. Angrist, through his immensely popular textbooks (co-authored with Steve Pischke) and short videos in the Marginal Revolution University series on ‘Mastering Econometrics’ (as Master Joshway), is an excellent communicator. He starts with a real-world problem and takes us through techniques of analysis, thereby inverting the approach of the standard econometrics textbooks that begin with dry, math-heavy abstract proofs, losing students along the way by the time the chapters end with examples.
In the age of big data and machine learning, are econometric techniques of causal inference increasingly redundant? Prof. Angrist argues that while data science helps ‘curve fitting’, i.e. it illustrates a pattern, it does not provide insights into causation. In other words, it neither enables an understanding of why we see a particular pattern, nor does it allow us to evaluate counterfactual scenarios. For that, we need econometrics, which does not rely on big data but on innovative ways of analysing data. Prof. Angrist believes that econometrics will continue to be relevant regardless of any advances in data sciences.
The work of the three Nobel Laureates demonstrates the immense power of good, rigorous empirical work and reminds us that rigour need not be seen in opposition to relevance, and that careful analysis can successfully challenge existing orthodoxy.