Page 38 - IJPS-3-2

Desta CG

2.3 Analytical strategies

The causal effect of fertility on the economic wellbeing of children is complicated by their endogeneity. Although there
are a few studies which failed to find endogeneity (Orbeta, 2005), the fact that fertility is endogenous to maternal work
participation is widely acknowledged in the economic demographic literature, in the presence of which the use of the
ordinary least squares estimator biases the effect of the number of children.
While the econometric literature offers various approaches to account for endogeneity, one of these is the use of an
instrumental variable. Using instrumental variable methods yields unbiased estimates even when fertility is or is not
exogenous (Schultz, 2007). Different studies used different instrumental variables to generate exogenous variation in
fertility. These include, for example, twin first birth (Chun and Oh, 2002; Kim, Engelhardt, Prskawetz, et al., 2009;
Rosenzweig and Wolpin, 1980a), abortion legislation (Bloom, Canning, Fink, et al., 2009), contraceptive choice of
couples (Kim and Aassve, 2006), sibling sex composition (Angrist and Evans, 1998; Cruces and Galiani, 2007), sibling
sex composition and contraception unavailability (Aassve and Arpino, 2007) and sex of the first birth (Chun and Oh,
2002; Orbeta, 2005).
The present study uses two-step instrumental variable probit (ivprobit hereafter) method, which is one of the most
common instrumental variable estimators (Wooldridge, 2009). The instrumental variable used consists of sex
composition of the first two siblings born to a mother (same sex = 1; otherwise, 0). This instrument is chosen because
sex composition of children is a random assignment and hence the sex of the siblings has no direct significant effect on
maternal participation in economic activities while it impacts the number of children.
In this procedure, the first step equation uses ordinary least squares to predict the number of children as a function of
the sex mix of the first and the second siblings, controlling for other covariates. Once the number of children is
exogenously predicted in this way, the final equation which estimates the mother’s work participation can be specified
by inserting the predicted number of children as key independent variable of interest, also controlling for the same
covariates in the first equation (refer to Appendix A for details).
3 Results

3.1 Characteristics of the study population
The tables below offer some descriptive statistics on the demographic and economic characteristics of sample
households. Table 1 and Table 2 provide mean values and frequency respectively for sample households on selected
demographic and economic variables across the rural-urban economies. As expected, Table 1 shows that households in
the urban sub-sample have higher average age at first marriage/child bearing and years of schooling compared to
households in the rural sub-sample. Table 1 also shows that household members including children for the urban sub-
sample work for longer hours (perhaps due to urban children’s older average ages) compared to their rural counter parts.

Table 1. Demographic and economic characteristics of sample households (means)
Mean values
Variables
Full sample Urban sub-sample Rural sub-sample
45.8659 48.5623 41.9875
Age of household head (years)
(11.9563) (12.3264) (9.8497)
5.5412
4.8911
4.3952
Number of children of the participant (2.0945) (2.1098) (2.4102)
13.7410 17.0014 11.0108
Average age of children of the participants (years)
(8.0197) (8.9971) (5.8945)
14.5276
16.9961
15.7618
Age of the participant at first marriage (years) (3.61805) (3.85991) (2.88352)
17.3415 18.7034 16.9856
Age of the participant at bearing first child (years)
(3.0002) (3.5142) (2.4315)
Participant’s years of schooling* 3.7981 5.9107 2.0098
(4.8475) (4.8910) (2.7458)
15152.3254
11001.7491
13904.0397
Value of household assets (ETB)** (15047.4125) (16049.1152) (15012.3124)
3.02 4.00 2.1
Mean hours of work per day by household members (excluding parents)
(0.21) (0.34) (0.21)
N 493 248 245
Note: Standard deviations are reported in parenthesis. Source: Survey data (2010 and 2013).
* if no formal education attended, years of schooling is recorded as 0.
** 1 USD=16.636 ETB on December 2010, and 19.1218 ETB on 31 December 2013.
32 International Journal of Population Studies | 2017, Volume 3, Issue 2