Potential Analysis Topics

[amichaelsen]

This issue is for brainstorming possible analysis topics.

[amichaelsen]

Data Collection/Processing

• Try getting original PUMS data and processing
  • Identify which dataset (3 year? 1 year?)
  • State zips vs table generation?
  • Identify which columns if table generation
  • Convert R script to python
• Get newer data from PUMS in the same way to make time series comparisons

Reproduction

• Reproduce much of the analysis from the FiveThiryEight article

New Analysis

• Compare popularity of majors/major categories
• Look at the percentage of women in various majors/major categories within STEM

[Stephenouu]

Some of the topics while looking at the data dictionary

1. Comparing the different popularity/all of majors category employment/unemployment rate over time if we can extract the dataset from recent years. Or we can just compare the rate between 2010 and 2012.
2. Higher level of education means better life quality (since there are variables about the water bills, fuel, transportation to work, medical care for low-income family, food stamp service) - data from PUMS
3. which major would bring the best life quality?
4. salary of one major change over 10 years? (use the 2010 data and 2020 data to compare)

[amichaelsen]

For comparisons over time there are two options:

• For the years being compared, get the recent grads data from each PUMS datasets and compare
• From the later pums dataset, split by age based on typical historical graduation dates and compare

[amichaelsen]

Questions:

• Which majors/major categories are most likely to be followed by a graduate degree?
• How do earnings by degree vary by race or citizenship?

Note: The FiveThiryEight analysis excludes people whose total earnings (not income) was negative. This seems to mostly include people with low income who are categorized as "self employed". I'm not sure if those people should be excluded in our analysis, we could try both or discuss which option is more appropriate.

Also, some fields when considered in a single year have very few graduates (e.g. petroleum engineering!) and these tend to give high variance (seen as both the highest and lowest earning groups). We could either group majors (by major category or a smaller subset) or eliminate majors with less than 10 (or some other cut off) people in the sample (note: 75% of majors from 2018 had at least 33 people, so 10+ should capture most majors)

[amichaelsen]

Other articles dealing with data:

• NYT - Is college worth it?
• What's it worth? The Economic Value of a College Degree.
• FiveThirtyEight - Women in lower paying STEM jobs