Attainment of Sex Preference in India
Examining the relationship between several demographic variables and the attainment of expressed ideal sex ratio.
Images from across Delhi, India
Introduction
With roots dating back centuries, son preference has been present in India and continued to the current day with implications such as millions of girls missing from the population. The prevalence of son preference stems from patriarchal practices such as a patrilineal inheritance an aging parent's dependency on their sons. Daughters can be seen as a burden, for reasons such as the practice of dowry payments when women get married and some women not bringing an economic contribution to their family. This can differ by region, as in rural Northern areas, women do not participate in agricultural production, but in rural Southern areas where rice-growing is practiced, women's involvement is crucial.
As India continues to grow in population, with predictions that it will be the most populous country by 2050, it is increasingly important to seek the end of son preference. Tools such as sex-selective abortions have grown in popularity in order to limit the number of daughters, and many daughters that are born into larger families get less resources than their brothers.
There are many studies that look into son preference in India, often by analyzing parity progression, the likelihood that families will try for more children based on the sex composition of their existing children. This study is different than many of the other sex preference studies in India. Instead of looking at parities, this study compares a woman's expressed ideal sex ratio with the actual sex ratio of her children. By looking at attainment of sex preference, we can see what variables might influence hitting the ratio or not. One finding that might come from looking at this type of analysis, is seeing if higher levels of wealth are correlated with hitting the desired ratio, which could give hint to sex-selective abortions taking place.
Map of India
Data
The data that is used in the analysis comes from the IPUMS DHS Global Health India 2015 dataset. More specifically, the data used is the Women survey, with responses from women ages 15-49, which is considered childbearing age. The variables that are used as predictors are education level, wealth status, urban or rural, and age. Several variables were used to create the outcome variable used in this analysis, including the three ideal number of children variables and variables indicating a woman's current number of children.
The outcome variable is an indicator of whether a woman's child composition met the ideal ratio that she expressed in the survey. This variable has three categories of values, positive numbers, negative numbers, or zero. A positive value indicates that the woman had more sons than her ideal ratio. A negative value indicates that the woman had fewer sons than her ideal ratio. A value of zero indicates that a woman has attained her expressed ideal ratio. The visuals below, show the breakdown of how this variable was created.
Conceptual Visual of Outcome Variable
Demographics of Gender Preference Subgroups
When performing this analysis, the original dataset was split into three subsets, based on the value for the ideal composition. Entries with a negative number made up the Daughter Preference Set, zero went to the No Preference Set, and those with a positive number made up the Son Preference set. Below are all visualizations of various demographic variables for each of these groups.
Split of Gender Preference Subsets
In total, there are 699,686 entries used in this analysis. The majority of entries showed no gender preference, at 76 percent of the data, with a gender preference, either son or daughter, being 24 percent of the data. The percentage of son preference, 20 percent, is five times higher than daughter preference, 4 percent, which suggests that son preference is still prevalent.
Education
The education levels on the right, from 0-3 represent, no education, primary, secondary, and higher, respectively. This bar plot gives interesting information about the various subsets that the data is split into. The daughter preference subset makes up a similar percentage of the total in all of the education levels.
The son preference subset has a higher percentage of the total for the lower education levels, around 30% for 0 and 1, and decreases to approximately 15% for 2 and 3. This suggests that there is a higher concentration of son preference amongst lower levels of education. This aligns with many other papers regarding gender preference in India, as lower education levels may be more connected to traditional views on children as discussed earlier.
The inverse of the son preference pattern can be found in the no preference subset. As the education level increases, so does the no preference subset's percentage of the total. This also does not come as a surprise, as higher levels of education can mean more exposure to movements for equality and different ideas regarding child composition.
Urban vs Rural
The composition of the urban and the rural categories are fairly similar, although the son preference group does make up a higher percentage of rural than urban. This again is not a surprise, similarly to lower levels of education, rural areas are expected to have a more prevalent presence of son preference than urban areas. This can be attributed again to stronger ties to traditional ideals and also a stronger agricultural presence.
Wealth
The wealth levels used in this bar graphs range from 1 to 5 and represent poorest, poorer, middle, richer, and richest.
Unlike the the two previous charts, daughter preference has a higher percentage in the higher wealth levels. Although it doesn't look like much due to the size of the daughter preference subset being much smaller than the others, there is a notable change from around 2.7% in level 1 to 4.3% in level 4.
The son preference's percentage decreases as the wealth level increases, while no preference increases. This pattern is directly connected to both the education level and urban/rural patterns. Rural areas with lower levels of education are likely to be situated in a lower wealth level.
Analysis
In order to look at how various demographic variables impact the attainment of one's ideal gender composition, I used several logistic regressions. The outcome variable was recoded so that not attaining one's ideal ratio is 1 and attaining the ideal ratio is 0. Below are the results of the regressions that are grouped based on the education and wealth variables.
Wealth Coefficient Plot
The plot to the right shows a split in the pattern as wealth levels increase. The daughter preference subset shows that as wealth increases, the likelihood of achieving the ideal ratio actually decreases. The two other subsets, son preference and no preference have the opposite trend, as wealth level increases so does the likelihood of attaining the ideal ratio.
Because son preference and the no preference subsets make up a large majority of the data, at 96 percent, we can say that the majority of the data shows that as wealth levels increase so does the likelihood of achieving ones ideal ratio. It is interesting though, that the daughter preference group is the opposite pattern.
Education Coefficient Plot
The results for these regressions do not follow as clear of a pattern as the wealth plot did. Similarly, there is still a trend between the different subsets. The daughter preference subset again shows that as the education level increases, the likelihood of achieving ones ideal ratio decreases.
The son and no preference subsets are not as similar in their results for education, with the son preference group not having much of a difference of coefficients for each education level. The no preference group, on the other hand, has an increased likelihood of achieving the ideal ratio as education levels increase.
Discussion
The coefficient plots above provide interesting findings relating to son preference in India. Although the daughter preference group showed a quite unexpected pattern in outputs, the majority of the data aligned with the expected output. These findings provide the support that the higher the education level and wealth status that an individual has, the more likely they are to reach their ideal composition, as they have more resources at their disposal. Women with lower education and wealth levels may not be able to use the same resources and may end up with a larger amount of children in trying to reach their ideal.
It is hard to say exactly why those in higher education and wealth levels have a better chance of attaining their ideal ratio. Perhaps they are content with the composition that they have and put that as their ideal ratio. It is also possible that better access to family planning or even sex-selective abortions play a role in the higher chance of attainment. Because of this, it is not possible to say with certainty what this study concretely shows us. This is not to say that it does not have an impact though.
By knowing if individuals are taking steps to attain their ideal compositions, we might have a better idea of how much of an impact gender preference still has today. It shows us where the issue is still prevalent and gives clues to different ways it might be expressed. This study and the many others that examine this issue of son preference are needed in order to take steps towards change.
References
Garg, S, & Nath, A. (2008). Female feticide in India: Issues and concerns. Journal of Postgraduate
Medicine, 54(4), 276-279.
Fred Arnold, Sunita Kishor, & T. K. Roy. (2002). Sex-Selective Abortions in India. Population and
Development Review, 28(4), 759-785.
Kishor, S. (1993). "May God Give Sons to All": Gender and Child Mortality in India. American
Sociological Review, 58(2), 247-265.
P. N. Mari Bhat, & A. J. Francis Zavier. (2003). Fertility Decline and Gender Bias in Northern India.
Demography, 40(4), 637-657.
Robert D. Retherford, & Tarun K. Roy. (2004). Effects of family composition on sex-selective abortion in
Punjab, India. Genus, 60(3/4), 71-97.
