*ORDER THE PAPER N**OW*

*ORDER THE PAPER N*

*OW*

You are interested in the inter-generational transmission of disadvantage. You would like

to know whether parental education helps improve the survival/health outcomes of their

offspring.

This is 2015 data from Madagascar (with some “tweaks” of the real dataset to suit our

purposes). The Millenium Development Goals (MDGs), which were outlined around

1990, listed “Universal Primary Education” as one of the key goals. In response to the

MDGs, the United Nations and other funding organisations funded a massive increase in

school construction in Madagascar during the late 1990s and early 2000s.

1. What is the correlation between parental education and the health of the child?

Estimate the correlation for each parent.

2. Run an OLS regression explaining the child’s health (Y) using mother’s education

and any other variables you choose to include. Interpret your results. You can

measure the child’s health with the variable “zscore”.

• What other variables are you including in the regression?

• Is the coefficient positive or negative?

• Looking at the standard errors or t-statistics, is the coefficient significant?

3. What selection problem are you concerned about, in the above regression?

4. Run an Instrumental Variables regression to capture the causal effect of

mother’s education on offspring health.

• Explain what regression you chose to run, and why.

• Compare your results with the results in question (2).

5. Critique the choice of instruments.

• Are there any problems with the instruments available?

• Are there any limitations on how we can interpret the results?

DATA: Each line of the database corresponds to a household. In each household, data is

collected for one child between the ages of 1 and 5, chosen at random (from all children

in that age group).

Variable Name Definition

Male Male = 1 if child is male, = 0 if child is female

AGE Age of person in years

Dirtfloor dirtfloor=1 if the floor is dirt, 0 if paved.

wcNONE wcNONE= 1 if there is no toilet, 0 if there is a toilet

Rich Long-term wealth, relative to the other households in the survey.

(So if it is negative, the child is from a poorer family.)

Usually independent of current income fluctuations.

religion Household head’s religion (coded as numbers)

ethnie Household head’s ethnicity (coded as numbers)

Newidmen Household identification number; you can ignore this variable.

Healthctr Whether birthplace had a health centre

Primsch Number of primary schools in district where you were born

Midsch Dummy variable for middle schools in district where you were

born: “1” if there are any middle schools, “0” if there are none.

Firstborn Is this the first born child?

Annedu Number of years of education

write_mlg Write Malagasy, one of Madagascar’s official languages

Zscore standardised mid-upper arm circumference (It’s a measure of

health: a higher number means no malnutrition. Negative means

below average.)

The same variables are repeated for the child’s mother: Then the variables have “m” in

front of them. So m_write_mlg is a variable that measures whether the mother can write

Malagasy.

The same variables are repeated for the mother’s partner (who is probably the father):

Then the variables have “conj” in front of them. So conj_write_mlg is a variable that

measures whether the father can write Malagasy.

Some variables are missing for logical reasons (for example, a child that is 5 years old or

less is not attending school, and most do not know how to write). Others have not been

gathered completely by the survey team. You will have to decide what to do about that,

and explain.

*ORDER THE PAPER N**OW*

*ORDER THE PAPER N*

*OW*