Page 67 - IJPS-2-2

Ramesh Babu Kafle

riables is used for analyzing the risk of higher order births.

2.3 Analytical Approaches
The median age at first birth and the cumulative percentage of women who had their first birth by
a certain age are obtained for different cohorts of women based on the NDHS 2011 dataset. The cu-
mulative proportions of women transiting to the next higher order birth are obtained based on life
tables constructed by taking all women who are exposed to the risk of birth of that particular order.
Proportions for the transitions to the second, third, and fourth birth are computed from NDHS data-
sets in 2001, 2006, and 2011 to analyze the aggregate changes in the transition probabilities to higher
order births. The lengths of the second, third, fourth and fifth birth intervals and the risk of higher
order birth are further examined from the NDHS 2011 dataset. Details of the methodology are given
in the following paragraphs.
th
th
th
The interval between k birth and (k+1) birth is called (k+1) closed birth interval. If women had
th
th
not given (k+1) birth by the survey date, this (k+1) birth interval is called an open birth interval.
Analysis of closed birth intervals depicts the pace of childbearing but cannot reflect parity progres-
sion. For this, one should also take the open birth intervals into account. The life table technique
used in this paper is based on a combination of both closed and open birth intervals and considers
the cases of open birth intervals as censored cases in analysis. Life tables constructed by pooling
open and closed birth intervals and treating open birth intervals as censored cases are a better way of
analyzing family building process (Lee 1993; Srinivasan, Pandey and Rajaram, 1994). Transition
th
th
th
probabilities between k and (k+1) births and the k order open birth interval are computed with
this approach. This approach of analyzing birth intervals enables explanation of the fertility transi-
tion both in terms of the amount and the pace of decline. Changes in the median birth interval indi-
cate the pace of fertility change and changes in the ultimate proportion of women who had their
next child within a certain period (parity transition) indicates the amount of fertility change. The use
of this method has been well documented in prior work (Srinivasan, Pandey, and Rajaram, 1994).
Monthly life tables are constructed for a time period of 10 years following prior birth (a negligible
number of births occur after an interval of 10 years from previous birth). The calculations are made
for all currently married women who have married only once. In the analysis, all multiple births are
th
treated as a single birth. The conditional probability of giving k birth between time t i and t i+1 is giv-
d C
*
th
en by q = i with n = n − i where, n * i is the number of women exposed to k birth at the start
i
i
i
n i 2
th
of the interval (t i, t i+1); d i is the number of women who gave i birth in the same interval and C i is the
th
number of women who reached and terminated from the same interval without giving k birth. Then,
th
p = 1 q− i gives the conditional probability of not giving k birth in the interval (t i, t i+1). The product
i
th
Πp gives the proportions not giving k birth by the end of the interval (t i, t i+1) and S = (1Π) p− i
i
i
th
gives the probability that a woman gave k birth by the end of the interval (t i, t i+1). The proportions
th
of women who gave k birth by the end of 24 months, 60 months, and 120 months are tabulated;
the curve of these monthly proportions is drawn and the median birth intervals are also tabulated.
th
Here, median birth interval refers to the length of the k birth interval in months by which 50 percent
th
of women give k birth.
Cox Proportional Hazard Model (also called Cox regression) is used to analyze the determinants
of the risk of occurrence of higher order births. Controlling for a set of background variables, the
relative effect of a particular characteristic of women on the risk of next higher order birth is ex-
amined. The model is semi-parametric and can be used even when the underlying distribution of the
hazard rate is unknown (Retherford and Choe, 1993). The other benefits of using this model are that
it is flexible, can accommodate time varying covariates, and also takes censored data into account
(Tarling, 2009). In this model, the failure rate of an event (i.e., birth of next higher order in the
International Journal of Population Studies | 2016, Volume 2, Issue 2 61