Descriptive Statistics of samples by background characteristics.
Table 1 shows the distribution of nationally representative samples (1, 62,877 and 1, 69,412) before and after marriage prohibition act 2006, with background characteristics respectively. For place of residence, there is slightly change in population i.e. 0.7 percent decrease in urban population while 1.2 percent increase in rural population after act was implemented. In education, there is decrease in no-education from 32.2 percent to 15.9 percent and improvement in secondary and higher education from 43.4–55.1% and 8.8–17.7% respectively in population before and after act. For religion, there is decrease in Hindu population by 1.4 percent while increase in Muslim population by 1.7 percent after act was implemented. Furthermore, there was decrease in poorest population from 19.5 percent to 16.8 percent while increase in middle, richer and richest population after act. There were increase in mass-media uses among population from 79.3 percent to 83.0 percent and also there was increase in relation to husband prior to marriage since 1996 from 13.8 percent to 15.2 percent in the sample. Additionally, household head of age greater than and equals to 55 years had also increase remarkably from 23.5 percent to 38.3 percent after 2006.
Socio-demographic characteristics of Child Marriage
Socio-demographic characteristics of Child Marriage with background characteristics in India is depicted by Table 2, which shows that there is significant decline in early-child marriage after implemented prohibition of child-marriage act-2006 i.e., from 48.0–26.9% but still early-marriage is happening around 27% in India. For place of residence, early marriage was high in rural population (54.3%) while in urban population estimate of early marriage was 36.2% before prohibition of child marriage acts, 2006. After act was implemented, it played significant role in reduction of early- marriage it comes to rural (31.7%) and urban (17.3%) population. Increase in education is significantly associated with decrease in early-child marriage for instance, result shows that before prohibition of child marriage act (2006) child marriage ranges from 59.6–12.4% among no-education to highly educated women whereas, after act was implemented it comes 37.1–4.7% among no-education to highly educated respectively. Similarly, there is significant declines in all ethnical-group, early-child marriages among scheduled caste decreases from 54.5–29.9%, scheduled tribes 53.5–33.5% and others 45.6–24.8% after prohibition of child marriage restrain act, 2006. For religion, the early-child marriage before 2006 among Hindu (48.7%), Muslim (51.0%), Christian (30.3%) and others (32.8%). After 2006, Hindu (27.1%), Muslim (29.3%), Christian (16.5%) and others (17.0%) and for wealth, the early-child marriage before act was highly associated in poorest (62.3%), poorer (59.4%) and middle (53.5%) while richer and richest had comparably less associated i.e. (42.2%) and (24.6%) respectively. After act implemented, it decreased significantly among all wealth-quintile i.e. 43.3%, 37.9%, 29.4%, 19.4% and 9.0% respectively from poorest to richest quintile. Percentage of early-child marriage is low in mass-media exposure population in both before and after implementation of child marriage act. East and central regions had high association of early-child marriage before act i.e. 57.7% and 56.6% respectively and after implemented act still east region is highly association of early-marriage i.e. 39.0% and in central region was 25.2% after act. There is significant decline in early-child marriage among household head age-group greater than 55 years i.e. only 23.8% as after act is implemented in India.
Adjusted and Crude Odds-Ratio of factors associated with early marriage before and after Prohibition of Child Marriage Act Using Multiple Logistic Regression.
Table 3 depicts the crude and adjusted effects of early-child marriage before and after implementation of act using multiple logistic regression analysis. Secondary and higher educated women were having 5% (AOR: 0.95; CI: 0.92–0.99) and 78% (AOR: 0.22; CI: 0.20–0.23) lower chances of marrying before their legal marriage age respectively after the act was implemented in compare to illiterate. In religion, the odds of early-marriage were less likely in Muslims, Christian and others as compared to Hindu before and after act in both crude and adjusted effect. The early marriage was negatively associated with wealth-quintile as their wealth improves the odds of having early child marriages decreases in both time zone. Furthermore, women having any mass media or female headed household or older household head were significantly contributing in reduction of child marriage in compare to their counter parts in both the cases before and after act implementation. Before act implemented, the COR and AOR of early-marriage were less associated in north-east (0.40and 0.51, p = 0.00), south (0.44 and 0.58, p = 0.00), north (0.50 and 0.70, p = 0.00) and west (0.51 and 0.63, p = 0.00) as compared to central region of India but after act implemented the odds of early marriage were observed to be high in the eastern, western and northern part of India as compared to central region of India. Additionally, irrespective of the act implementation it was found that women with prior relationship with their husband also lead to higher chances of child marriage.
District level Prevalence of Early Marriage
Figure 2 indicates the percentage distribution of early marriage before (1996–2005) and after (2007-16) Prohibition of Child Marriage Act, 2006 in Indian districts. In the map, the percentage of marriages before 18 years categorized in five shades of colour. Darker the colour, higher the percentage of early-marriage among different district of India. During 1996–2005, the states like West Bengal, Rajasthan, Madhya-Pradesh, Uttar-Pradesh, Bihar and some parts of Andhra-Pradesh district had higher percentage of marriage before 18 years’ ranges (60.1–80.1%) before act. Similarly, after implementation of prohibition of child- marriage act only two district have more than 60% of percentage of marriage before 18 years.
Univariate and Bivariate Moran's I Statistics
Appendix table 1 and Table 4 indicate the univariate and bivariate Moran’s I for the dependent and independent variables. Table 4 shows the spatial dependence for the district level percentage of Meso scale Variables with early-marriage before and after Prohibition of Child Marriage Act. Bivariate Moran’s I value for rural was 0.76 and 0.72 before and after implemented act respectively which shows there is high spatial auto-correlation of early-marriages in the rural-districts of India. Similarly, for Hindu, there was 0.74 Moran’s I values before act was implemented and it was decreased to 0.69 after act implemented. No mass-media, prior relation to Husband had Moran I value ranges from 0.50 and 0.54 respectively after the act was implemented. Female household head, 1-4 household members and household head less than 55 years had values around 0.70 for early-marriage which means there was still high spatial auto-correlation of early-marriage even after act was implemented.
Similarly, from appendix table-1 indicates spatial dependence for the district using univariate Moran’s I statistics before and after implementation of act, 2006. Moran’s I value and Z-value for early-marriage ranges from (0.79, 31.09) and (0.73, 28.99) before and after act respectively which shows highly spatial auto-correlated. Similarly, for rural, Hindu, 1-4 household members and household head <55 years is more than 0.70 which mean it was highly spatial auto-correlated before act. After act implemented, value more than 0.70 only for rural and household head less than 55 years.
Bivariate LISA maps indicating the spatial distribution of different covariates of early marriage before and after child marriage prohibition act
Figure 3 shows the bivariate LISA cluster maps indicating the spatial clustering and outliers of different independent variables across the districts of India, Before and After Prohibition of Child Marriage Act. Map A1 indicates 174 hot-spot regions in rural area depicting high regional dependence of child marriage before the Child Marriage Act which includes regions of Uttar Pradesh, Bihar, Madhya Pradesh, West Bengal, Rajasthan, Andhra Pradesh, and Telangana. Map A2 shows 136 hot-spot regions indicating high regional dependence of child marriage after the Child Marriage Act which includes regions of Rajasthan, Madhya Pradesh, Jharkhand, Bihar, West Bengal and a few parts of Maharashtra and Telangana. Map B1 shows 163 hop-spot regions depicting high regional dependence of child marriage before the Child Marriage Act which includes regions of Uttar Pradesh, Bihar, Madhya Pradesh, Rajasthan, West Bengal, Andhra Pradesh, and Telangana. Map B2 indicates 98 hop-spot regions depicting high regional dependence of child marriage after the Child Marriage Act which includes a few parts of Rajasthan, Madhya Pradesh, Maharashtra, Bihar, Jharkhand, and West Bengal. Similar results were found in other figures (Map C1, C2, D1, D2, E1, E2, F1, F2, G1 and G2) which shows higher number of hotspots districts were accounted for early marriage before the act implementation and reduced significantly after act implementation.
Estimates of Ordinary Least Square and Spatial Error/Lag Model Regression Analysis
Further table 5, Ordinary Least Square regression analysis have also been used to show the adjusted coefficient of the correlates for early marriage before and after prohibition of Child Marriage Act in terms of coefficient, standard-error and p-value for district level meso-scale correlates for India. A positive coefficient indicates that as the value of the independent variable increases, the mean of the dependent variable also tends to increase. A negative coefficient suggests that as the independent variable increases, the dependent variable tends to decrease. Early marriage was significantly associated among rural had coefficient value 0.095 and it increased to 0.231 after act was implemented. No education had negative coefficient and significant association with early marriages i.e. -0.003 before act and -0.006 after act but not significant. Hindu, prior relation to husband, female household head, and 1-4 household members had significant p-value and its coefficient value increased after prohibition of child marriage act, 2006 having R-square value 0.9893 and 0.9838 before and after prohibited act respectively.
Table 6, Since the OLS confirmed spatial autocorrelation in its error term for the outcome variables, we further estimated Spatial Error Model (SEM) and Spatial lag model (SLM). The underlying assumption of a spatial lag model is that the observations of the dependent variable are affected in the neighbourhood areas whereas the spatial error model is used to consider the effect of those variables which are not present in the regression model but have an effect on the outcome variable. Rural, Hindu, prior relation to husband, female household head, 1-4 household member and household head less than 55 years was significantly spatially associated with early marriage before and after act in districts of India. In lag model, rural and household head less than 55 years had significant higher coefficient value i.e. 0.092 and 0.741 respectively. In error model rural (0.234), Hindu (0.072), prior relation to husband (0.021), female household head (0.053)1-4 household member (0.081), and household head less than 55 years (0.486) respectively.