Internal Code: MAS1692
Report Writing Assignment:
Bedard-Deschênes (BD) looks at the effect of divorce on the economic status of women. The simple regression of economic status on divorce is probably confounded by selection: the kinds of processes that lead to divorce may also affect a woman’s economic standing and labor force attachment. What you need is either a good experiment (unlikely in this case) or an instrumental variable. BD use the sex of the first child as an instrument. Based on the literature, a first-born daughter has been found to increase the probability of marital dissolution compared to the birth of a first-born son. Why is an interesting question – it’s worth reading the paper and some
of the references they cite.
Using micro census data for the US (IPUMS) they can figure out the gender of the first child, whether or not the mother remained married, and her economic status (as measured by income, labor supply, and other sources of income). Your task is to read the paper, think hard, and figure out how to do this using the same data.
Replication: the basics and some (free) hints
Your task is to replicate tables 1 to 5 in the paper. Although the goal of a replication is normally to reproduce exactly their results, my attempts to do this have gotten me close, but not to an exact replication. So don’t pull out your hair if you can’t. When you’re close, you’ll know it. It is helpful to divide the process of replication in three major steps:
(i) get the right samples;
(ii) define the right variables;
(iii) obtain the right descriptive statistics and run the right regressions.
Note that these three steps are consecutive: you will not get the right results if you are not using the right samples and the right variables.
So you will first need to obtain the raw data, and “clean” (redefine, adjust, reshape) the data. To obtain the raw data, go to the IPUMS web site, and create a user id and password. Then you can get to work selecting the data set and variables. How do you figure out what data set and variables to choose? For this you read the paper, all of it but especially Section 2. You use the description of the variable in the paper to search for the right variable name on the IPUMS web site.
While you’re at it this stage, also take a look at the detailed variable descriptions for some basic variables. For example, you’re certainly going to need sex. Searching for this variable on IPUMS, turns up “sex”. Click the “Codes” tab, and you’ll learn the key fact that 1 denotes male and 0 denotes female. Note that some of the variables may not follow the conventional coding, and in some cases it may be required to recode some of them in order to get a usable format.