Economics of Faith, Race, and Community: Reframing (June 13-20)

Turns out, I don’t have geographical information in the GSS, which is what Gruber used to predict religious attendance using religious market density. Gruber uses geographical data to compute religious market density to then predict religious attendance. He then uses that religious market density from the GSS to predict outcomes in IPUMS–he computes predicted religious market density, which uses the ancestral densities of the geographic locations as an instrumental variable for religious market density (predicted religious density).

Predicted religious density definition: “density of that ancestry in an area times the probability that the ancestry is of a given religion.”

 My main problem: I cannot compute religious or ancestral density in the GSS without the geographical location.

I started to brain storm other data sources that include data on religiosity and economic outcomes. Pew Research has a data set for the year of 2014 that includes data on religious attendance, state, race, etc. However, I am not sure that racial density is good enough to predict religious preference instead ancestral density. Generally, ancestral data can predict if a person is protestant or catholic.

Here is where I started to reframe my project and thought about the purpose of my research project. My goal was not necessarily to just replicate Gruber’s work; it’s to study how religious beliefs across all groups can inform their economic decision making – I wanted to look at more religions than just Catholic/Protestants. I want to look at the communities in America, their religious composition, and how that affects their daily lives. Does it have any effect at all?

The main difficulty in looking at these issues empirically is the availability of data. Here is what I can do: I can do a cross-section analysis with the Pew Research data. I have the state information, I have lots of information regarding their religion: preference, importance, social issues, attendance. I can find religious density at the state level. What I cannot do is replicate Gruber’s work exactly.

My plan for the Pew Research data is to recode the religious data into categories: Catholic, Protestant, Jewish, hindu, muslim, buddhist, atheist, etc. Then I will find the share of people that share the same religion in the state, and regress with religious attendance. I have income and education data as well.