Download presentation
Presentation is loading. Please wait.
Published byEthan Freeman Modified over 9 years ago
1
How family size affects children’s schooling and work in Mexico Emla Fitzsimons Bansi Malde 27 Feb 08
2
Introduction Question of Interest –Does family size affect children’s school and/or work participation? Theory –Quantity-Quality model: bigger families (quantity) reduce investment in schooling (quality) Empirical Issues –To test how well the quantity-quality model fits reality requires exogenous changes in fertility that are uncorrelated with preferences or budget constraints –We use Instrumental Variables to identify the causal effect
3
Instrumental Variables We use the following instruments –Twin births: increases family size by definition –First n children are of the same sex: increases family size if parents have preferences for having children of both sexes Note, the literature that considers the causal effects of family size basically always uses these instruments
4
A glance at the literature! Most of these studies consider developed countries and outcomes different to ours (female labour supply) Literature most comparable to what we do (effects of family size on education in developing countries): –India: Rosenzweig & Wolpin (1980) –China: Qian (2006), Rosenzweig & Zhang (2006) –Korea: Lee (2004) –Brazil: Ponczek & Souza (2007) –Colombia: Baez (2007) We will hopefully improve on them thanks to extremely large samples
5
Methodology Basic model IV first stage –Using same-sex instrument –Using twins instrument
6
Data The main source we use is the Mexican ENCASEH survey: cross-sectional census data collected across marginalised rural areas b/w 1996 and 1999 Info on individual, household and locality characteristics –Restrict sample to children aged 11-17 –Drop households with both parents not living together/not married eldest child>18 >1 household head So we’re left with a sample of ~600,000 households and ~1.1million children aged 11-17 Note, average # of children per family in our sample is 4.3 We’re also going to merge these data with the PROGRESA surveys, as there’s v useful info for our purposes in these - we’ll come back to this later
7
Randomisation check - same- sex of first 2 births
8
Randomisation check - twins at second birth
9
First stage results – Boys and Girls
10
First Stage Results - Girls
11
First Stage Results - Boys
12
TSLS estimates for First Born Boys and Girls
13
TSLS estimates for First Born Girls
14
TSLS estimates for First Born Boys
15
Issues/Next Steps Validity of Instruments Economies of scale? (Rosenzweig and Wolpin (2000))
16
Issues/Next Steps Robustness Checks –Compare results across different instruments (what we have just shown) –Use an instrument that should suffer much less from this issue: twins at second birth that are of different sex from first- born sensitivity first stage.doc; sensitivity second stage.docsensitivity first stage.docsensitivity second stage.doc –Use PROGRESA consumption data to see if there is any evidence of economies of scale we observe expenditure on children’s clothes and shoes, separately by sex: we are going to use these data to see if sex- sameness affects these expenditures we also observe value of assets, may be useful –Assess sensitivity of parameter estimates to different correlations b/w the IV and the error term using method of Ashley (2008); (note refine correlations using info from PROGRESA data)
17
Issues/Next Steps Results obtained from the IV are all LATE and affect only particular types of households –How do we reconcile all these different LATEs?
18
Any Comments/Questions?
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.