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

A+| A| A-

'Legislating' Social Change: Strange Case of the Sarda Act

This paper examines whether laws with inadequate penal provisions can still work towards eradicating undesirable social customs. Data on age at marriage from family genealogies in India from 1905 to 1980 is used to examine the effectiveness of the Child Marriage Restraint Act (1929). The findings reveal that though the law lacked adequate penal clauses, it still contributed towards increasing the age at marriage. It worked because a cultural shift towards later marriages was already taking place; the law worked by reinforcing the cultural change. The paper draws attention to the importance of the context in which social legislation operates and the contingent factors that determine its success or failure.

Special articles

‘Legislating’ Social Change: Strange Case of the Sarda Act

This paper examines whether laws with inadequate penal provisions can still work towards eradicating undesirable social customs. Data on age at marriage from family genealogies in India from 1905 to 1980 is used to examine the effectiveness of the Child Marriage Restraint Act (1929). The findings reveal that though the law lacked adequate penal clauses, it still contributed towards increasing the age at marriage. It worked because a cultural shift towards later marriages was already taking place; the law worked by reinforcing the cultural change. The paper draws attention to the importance of the context in which social legislation operates and the contingent factors that determine its success or failure.


I Introduction

here is good reason for a healthy dose of scepticism about the ability of laws to drive socially entrenched conventions into oblivion. Conventions are sticky because unilateral breaking away from them proves far too costly. Therefore, conventions may continue even when there is no particular rationale for their existence. Conventions like foot-binding in China, female genital mutilation in sub-Saharan Africa or child marriages in India, all strategies for control of women’s sexuality, arose in specific circumstances and continued to exist even when these conditions responsible for them no longer existed. It has been argued that female genital mutilation arose among slave traders [as a way of avoiding sexuality among female slaves] since virgin slaves brought more money. It became a marker of status and persists even among nuclear urban households in regions of sub-Saharan Africa. Abdalla(1984) surveyed 70 female and 40 male university students in Somalia and found that though 60 per cent women and 58 per cent men thought that female genital mutilation should be abolished, 66 per cent women and 50 per cent men still said that they would mutilate their daughters.

Mackie (1996) argues that the Chinese convention of footbinding emerged through increased resource polarisation and the desire of individuals for children which were biologically their own. Feet-binding was an attempt by wealthy polygamous individuals to minimise sexual philandering by their several wives, whereas status maximising households from lower socioeconomic strata evolved an interest in supplying women whose feet were bound. Gradually, conventions emerged which used the size of the foot as a social status marker: “Smaller the foot, better the family”. These conventions created belief traps where the affected see their inferior equilibrium as “normal”. Such belief traps are common in other contexts too. One of the effects of genital mutilation of women is to make urination difficult.

Lightfoot-Klien (1989) reports that to the question about whether urination was difficult, women respondents answered “normal”. Yet, when asked how long it took, the answer was “normal…about 15 minutes”. Such belief traps are typical of Pareto-inferior conventions.

How does civil society attempt to change these conventions? In particular, can legislation be effective at all in such cases? Have there been instances when societies have successfully legislated a convention into oblivion? If so, what can we learn from them? These are the questions that this paper attempts to answer.

In this paper, we analyse the effectiveness of the Child Marriage Restraint Act (1929). The final enactment of the bill raised the age of marriage for girls to 14. Rao Sahib Haribilas Sarda moved a bill in the Constituent Assembly on February 1, 1927. This bill, which aimed at restraining the solemnisation of child marriages came to be known as the Sarda bill. The process of enacting the bill, right from its initial conceptualisation till its final enforcement in April 1930 was marked by controversy. The bill was finally passed in the Constituent Assembly on September 20, 1929 by a large majority and became an Act on October 1 of the same year. It came into force on April 1, 1930. Some sections of the orthodox opinion saw the law as another instance of invasion entirely avoidable of Indian religious feelings by an alien government. Protesting the bill, the Bhala, an orthodox newspaper wrote “It is a question of the utmost importance whether the government has a right to interfere with a religious affair like marriage, which is one of the 16 sacraments of the Hindus. We strongly hold the view that an alien government should on no account be allowed to settle religious questions…If bills like this are passed by the legislative assembly, the Hindus will have to say that they lost their religious freedom in 1929, just as they lost their political freedom in the past”1. Not only Hindu, but Muslim orthodoxy was also opposed to the bill, which applied to all major communities in India.

On the other hand, there were others who believed that legal intervention was absolutely necessary in order to overcome

Economic and Political Weekly January 13, 2007


Economic and Political Weekly January 13, 2007

societal prejudices. Some even envisaged a special role for the law in this regard. For example, while commenting on the bill in the assembly, Srinivas Iyengar argued “Members of this legislature must create an opinion; and in the matter of this description, the opinion must go down from the legislature to the populace rather than from the populace to the legislature.”2 In fact, demand for such a law had arisen towards the end of the 19th century. At that time, only a handful of reformers demanded a law. However, by 1930, public opinion seems to have mostly favoured the bill as seened by the number of people and associations that supported it when the government circulated it for public opinion; 761 out of the 1,209, i e, 63 per cent witnesses examined by the Age of Consent Committee (1928-29) were in favour of raising the age at marriage by law.

The law as it came into force in 1930, provided for a fine of Rs 1,000 in addition to imprisonment up to one month for adults solemnising the marriage of a girl under 14. In case the groom was above 21 years of age, he too could be imprisoned. In spite of these provisions, there is reason to doubt the credibility of the penal clause. The fine could hardly be a deterrent if it was considered as just another expense to be incurred in connection with the marriage. It is unclear who would have complained to the authorities against such a marriage. Only complaints lodged within one year of the marriage alone were valid. The complainant would have to execute a bond of Rs 100 to compensate if the complaint turned out to be false. Women, even if convicted, were not to be imprisoned. Adults solemnising such a marriage could always argue that they were “under the impression” that the girl was over legal age. In a system with inadequate documentation to prove the actual age, such a claim could easily see such adults through. Thus, the punishment stipulated under the law could hardly be considered a credible deterrent for a determined parent. As a result, the law is widely thought to have failed in its objective. The 1931 Census report indicated that the act was a “dead letter” in most provinces [Rathbone 1934]. In August 1932, two and a half years after the Act was passed, there were only 173 prosecutions, of which 167 (more than 95 per cent) were convicted. Even when prosecution was likely to succeed, the numbers actually prosecuted seem to have been small. This is evidently the pattern even today. The number of cases brought to court for all India as a whole was 85 in 2001, 113 in 2002 and 63 in 2003. These are minuscule numbers given that the practice of child marriage is very widespread in large parts of India. On the basis of figures such as these, modern evaluations of the Age at Marriage Restraint Act (1929) are highly pessimistic. Some even argue that the age at marriage actually went up in the six months before the passing of the Act [Whitehead 1995].

This paper examines afresh the issue of effectiveness of this Act. Section I gives the data and the empirical findings, indicating that contrary to scholarly perception the Act seems to have succeeded in raising the age at marriage at least in one situation where adequate data do exist to take a statistically sound view of the problem; section II tries explain why the law would have been successful in spite of the difficulty of persecution.

The data that we have used have been collected from various Kulvritantas published by the Maharashtrian brahmin clans over the period 1940 to 1994. These Kulvritantas are available in printed form and are easily accessed.3 The Kulavritantas are family genealogies mainly concerned with the male lineage. However, some of them do provide us information on women. Typically, the following information on any given woman is to be found: (a) the year of birth; (b) the marriage year; (c) level of education if any; (d) the dates of birth of male and female children; (e) the birth date of the husband, and (f) father’s and husband’s occupations and addresses. Our sample is not even remotely representative of the entire gamut of Indian womanhood. It is important to keep the salient features of our sample in mind. The households considered in this sample have a much larger proportion of urban households in comparison to the all India average. The literacy level too is much higher. For example, in 1926-27, in the Bombay Presidency, 9.2 per cent of “advanced Hindu” girls were receiving some instruction in various communities, as against 1.4 per cent for intermediate and 0.8 per cent for backward Hindu communities. In 1931-32, these proportions had reached 11.1 per cent, 1.7 per cent and 1.0 per cent respectively. Though lower than Parsis, Europeans and Anglo-Indians, the numbers for “advanced Hindu castes” indicate the unrepresentativeness of our sample as far as women in general are concerned [Government of India 1932:159]. There are clear economic status differences too. In the sub-sample of rural households, agriculturist households are typically landowners and there are no landless agricultural workers in the sample. The other rural occupations of husbands and fathers are mostly restricted to teaching, priesthood and shop-keeping. The urban occupational structure is more diverse, with service occupations being predominant, though not all are necessarily white collar jobs.It is obvious that our sample is not representative of the all India population.

The fact that our sample only consists of women belonging to a specific upper caste by itself is less damaging than appears at first brush. Historians of India are beginning to appreciate the fact that apart from caste, Indians have various other identities, and their responses are determined by a complex interplay of these identities with material conditions[see Chandavarkar 1994]. Our data do have a wide variety of other identities and material conditions, with the limitations mentioned in the previous paragraphs. It is to be doubted whether caste can be considered a single overwhelming explanatory factor, neglecting this spectrum. It is extremely doubtful if people’s behaviour can be entirely predicted if their castes only are known. Hence, neglecting only the caste aspect need not destroy an argument in its entirety.

We have collected data on 6,200 women born between 1818 and 1960. In this study, we are concentrating on the 2,566 women whose year of marriage was t for t = 1900,1901…1980 and for whom the birth year as well as the year of marriage had been recorded. Thus, we have an average of 31 sample points for each year. But the distribution across time is rather uneven. Appendix 1 gives the estimated values and the number of observations for each year. Appendix 1 also gives details of the method used by us to estimate the mean age at marriage for each

t. We have assumed that the time till marriage is a random variable that follows the Weibull distribution. This distribution is commonly used to model life time data. The important feature of this distribution is that the hazard rate, which is the conditional probability that an event will occur at time t for the first time, is not constant but increases with t [see Cameron and Trivedi 2005: 577 and Rohatgi 2003: 407]. This gives it a distinct advantage over other distributions like the exponential that are sometimes used to model the hazard rate. This is also the reason why we have not used the simple arithmetic average of women marrying in a particular year. When modelling age at marriage, explicit account must be taken of the dependence of the

Economic and Political Weekly January 13, 2007 probability of getting married at a given time on the time elapsed till marriage. The Weibull distribution allows us to do this. We have computed the Weibull parameters for each of the years from 1905 to 1980 and then used these parameters to compute the mean age at marriage for each of the years. Since the number of observations for some years before 1905 was very small, we have used the data from 1905 to estimate the parameters. Yet, the fact remains that even after 1905, some years have a rather small number of observations as can be seen from Appendix 1.

The steady increase in the estimated age at marriage can be directly seen from Figure 1. Our aim is to test if the enactment of the law hastened the process beyond the rate at which the change was already occurring. To do this, we estimate the following regression:

ln (age(t)) = α + δ* dummy + k*t +η*t* dummy

+μ*ln(age(t-1)+ ε(t)) ...(1) where ln(age (t)) is the natural logarithm of the estimated age at marriage at time t, dummy is a variable which takes values 0 from 1905 to 1929 (the year in which the bill was brought before the Constituent Assembly ) and 1 thereafter, t is a time index stretching from 1905 to 1980, and ε(t) is the error term. The lagged dependent term has been introduced to take care of auto-correlation. The null hypothesis of no impact of the law over and above the secular change already taking place corresponds to the following restrictions on the estimated parameters:

δ = η = 0 (2)

We report below the estimated regression: ln(age(t)) = 1.4217+0.10823*dummy+0.0064818*t

(6.130) (3.864) (4.657) –0.002 *t*dummy+0.41511*ln(age(t-1)) (–1.744) (4.40) R-bar square = 0.978

Figures in brackets are the estimated t values. All the estimated coefficients are significant at 1 per cent, except the coefficient on the interactive dummy which is only significant at 10 per cent.

The restriction implied by the null hypothesis was strongly rejected. The associated F value of the test was 8.75, which was significant at 1 per cent with 2 and 70 degrees of freedom. Since the variance of the estimated mean age depends upon the number of observations used to estimate the parameters, the above regression parameters have to be adjusted for heteroskedasticity (see Appendix 1). The estimated t values therefore are based on Whites’ heteroskedasticity consistent variance – covariance matrices.

The coefficient on the interactive term is negative, which would seem to be surprising. However if one takes account of the fact that higher the age at marriage, the subsequent growth is likely to be comparatively slower (specially given that the age at marriage seems to have a terminal bound somewhere in the early 1920s), the negative coefficient would seem to be less surprising. It should also be noted that this coefficient is significant only at 10 per cent.

The Sarda bill sought to prevent marriages of girls under 14 years of age. It might be instructive to see how the proportion of girls who got married under 14 years of age to the total marrying in a given year was changing with time. This proportion, shown in Figure 2, is highly relevant to our study because the law only applied to marriages under 14 years. It was declining in the period before the Act came into force, but the decline seems to have much sharper towards the end of the 1920s, precisely at the time when the discussion about the Act was featuring prominently in the public sphere.

We might be interested in testing whether the percentage of girls who married below 14 years of age declined after the Act came into force. This can be done by replacing the dependent variable in equation (2) with the proportion of girls marrying under the age of fourteen every year, which we will call propor(t). The restriction (2) will continue to apply as the maintained null hypothesis. In this case too, the t values are based on heteroskedasticity consistent variance covariance matrix. The following is the estimated regression equation: Propor(t) = 0.59519 + 0.38159*propor(t-1)

-0.01499*t -0.5261*dummy + 0.01394*t*dummy (3)

Adjusted R-square = 0.93, all the variables except the interactive dummy, are significant at 1 per cent.

This equation implies that the proportion of girls marrying before reaching 14 years of age in our sample became almost zero after the implementation of the law. Again, restriction 3 is not accepted, the F statistic turning out to be 8.6, significant at 1 per cent.

The above results do indicate that we can safely reject the null hypothesis that implies no change in the age at marriage before and after the law. Without confounding statistical significance with causation, we can safely say that the law seems to have worked. It can also be said that modern perceptions of the Act of 1929 are unduly pessimistic.


We saw that the penal propositions of the Child Marriage Restraint Act, as it came into force in 1930 were not very enforceable. Breaches of legislation were not observable publicly because no parties to the transaction had an incentive to complain to the authorities. In addition, proving guilt in a court of law could have proved difficult because the recording systems might be poor. In spite of this, the Act seems to have worked.

Economic and Political Weekly January 13, 2007

To read the full text Login

Get instant access

New 3 Month Subscription
to Digital Archives at

₹826for India

$50for overseas users


(-) 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