eduViz is a visualization project that aims to make sense of the recently-released College Scorecard Data. The dataset, provided by the Department of Education, contains information on higher-education institutions. Some of the metrics provided are admission rates, student completion rates, debt and repayment, earnings, demographics and more. The dataset contains nearly twenty years of data, with about 1700 variables tracked. Most of this visualization will focus on data from 2013.

Availability of variables and their associated values becomes sparse in earlier years of the dataset. Cohort earnings data is from 2011 (most recent available).

This analysis is designed for data scientists, analysts, and others with statistical literacy. Others looking to create data graphics, articles, or more focused analyses will find this project useful. In particular, members of the Kaggle community can benefit from it. An effort is made to ensure that the analysis is replicable and transparent for others. While analysis is thorough and focuses on many variables, it is by no means exhaustive.

Process - Data manipulation

Due to the immensity and complexity of the dataset, a meticulous approach to data manipulation is required. In the interest of providing full transparency to the process, I have documented every step taken to parse the dataset. This is also done to ensure that other data scientists and developers can reproduce the parsed data and get the same result. The key tools used are Python, sed, and csvkit. A detailed Jupyter Notebook with code in Python and Bash is available.

View Notebook

Before diving into the data, it is useful to understand certain terms as they relate to higher education. It is particularly important to understand terms related to lending practices. Recently, many institutions' lending practices have been called into question, and much controversy arises on the subjects of tuition, loans and grants.

• Title IV

  Most data from the College Scorecard dataset apply only to Title IV students, or students receiving federal grants or loans. EduViz focuses moreso on grants, a type of financial aid that doesn't need to be repaid. Title IV aid varies in the conditions necessary to receive it - often it is provided on the basis of a student's financial need, or outstanding academic achievements. For the purpose of this project, analysis will focus on the most relevant form of Title IV aid, the Pell Grant.

• The Pell Grant

  A Pell grant is aid given by the federal government, based on student need. A Pell grant is an entitlement, meaning that it does not have to be paid back by the borrower. Our dataset provides the percentage of students receiving Pell grants for each given university. This can be used as an indicator of how well an institution serves low-income students. Institutions with a high percentage of students receiving Pell grants will better accomodate poorer students.

  Pell grants are administered by the US Department of Education, who assess information reported on a student's FAFSA (Free Application for Federal Student Aid.) Metrics such as expected family income are used to determine a student's eligibility. Pell grants are generally considered the foundation of a student's financial aid package. These visualizations may use the variable PctPell to describe the percent of students receiving Pell grants at an institution.

• Public, Private, and for-Profit Schools

  One of the focuses of this visualization project is to understand differences between public and private universities. The key difference is that public universities are funded by state governments, whereas private universities don't receive any state funding. Public universities generally offer lower tuition rates because they are subsidized, while private universities rely on higher tuition rates and generous private contributions to cover operating costs.

  Private schools exist as either nonprofit or for-profit institutions. For-profit schools are operated privately by profit-seeking businesses. While for-profit schools are thought to fulfill an economic need of education, they are frequently criticized for their lending practices, tuition costs, and dismal student outcomes. In the dataset, the type of institution may be referred to as Control.

• Accreditation

  An accredited school is one recognized as meeting standards of education, admission, and student outcomes. The aim is to ensure that schools maintain a quality level of education. Accrediting agencies exist to determine how well certain schools are meeting their standards; these external orgnizations will determine whether or not a school is given its 'accredited' status.

• Public schools are well-funded and generally yield favorable outcomes.

  Evaluating if an education is worthwhile can be modeled using cost and earnings data. Each school is given an earnings-to-cost ratio, calculated by dividing median annual earnings by annual cost.

  In plotting this metric, an interesting trend occurs for public schools. The histogram shows that almost every public school has an earnings-to-cost ratio greater than one.

  The key inference of this observation is that public schools appear to be well-funded, offering affordable opportunities to enhance occupational prospects.

  
• Schools with greater share of ethnic minorities generally have lower completion rates and earnings.

  The scatter plot below shows schools by completion and admission rates. Each dot is a school; horizontal lines indicate average completion rates. The darker the color of the data point, the greater the proportion of black students at that university.

  

  It is interesting to note that for public schools, along with quite a number of private for-profit schools, those schools with greater proportion of black students tend to have lower completion rates. It is difficult to interpret this observation, but one can speculate.

  There is a demonstrable correlation between socioeconomic status and academic ability. A recent New York Times article shows the degree to which income and affluence affect academic achievement. The College Scorecard analysis affirms the notion that students from poorer families will experience greater challenges academically, and that academic gaps persist into post-secondary education.

• A high-cost of education does not necessarily guarantee good earnings.

  Private schools that focus on providing educations in art, music, or other non-STEM degrees can provide dismally-low earnings compared to their cost.

  

  There are an alarming number of expensive private schools that report low median earnings for their students (10 years after matriculation). The table below lists the 10 most expensive schools for which median earnings are under $25,000:

  ---

  Type	School Name	Median Yearly Cost	Median 10-year Earnings
  Private nonprofit	Landmark College	64,233	22,200
  Private nonprofit	San Francisco Art Institute	55,809	24,600
  Private nonprofit	Manhattan School of Music	53,903	22,700
  Private nonprofit	Pennsylvania Academy of the Fine Arts	43,656	20,900
  Private nonprofit	Maine College of Art	43,313	23,700
  Private nonprofit	Pacific Northwest College of Art	43,235	22,600
  Private nonprofit	Naropa University	42,575	18,600
  Private nonprofit	Montserrat College of Art	37,866	25,100
  Private for-profit	Advertising Art Educational Services DBA School of Advertising Art	37,702	22,200
  Private for-profit	AmeriTech College-Draper	37,574	23,800

  ---

  Important to note is that Landmark College is an institution that specializes in educating students with learning disabilities. This school differs greatly from most in its objective and can be considered an anomaly.

While analysis focuses on aggregate metrics from this dataset, some will take interest in viewing a given school.
Use the Search for a School form to find a school, select it and click View summary.

In order to hold an institution accountable for lending practices, and maintaining standards of accessibility and affordability, it is helpful to understand where universities get their money. These diagrams detail tuition revenue per full-time student and average monthly faculty salary. These metrics provide a proxy for how universities source their revenue, and the caliber of teaching faculty they strive to maintain.

---

Cost and Earnings

Comparing cost and earnings is a good place to start. While there are some limitations with the earnings data, the aggregate figures can give us a good insight as to overall market trends.

---

The dataset gives us a generous array of variables with which to examine demographics. With these variables, we can paint a broad picture of what higher-education student population looks like at a number of institutions. The intricacy and granularity of our variables can also allow us to drill down further and look for patterns of interest.

Some of the key demographic variables we are interested in are:

---

Race: High-level Overview

---

Race: Admission and Completion

Many of the most prestigious and costly universities also have high completion rates. However, these schools are often very homogenous (Caucasian) and lack racial diversity. One finding of this data exploration is that schools with greater diversity have lower completion rates. It isn't feasible to draw a causal link with this information, but lower completion rates in more diverse schools suggest that ethnic minorities may not have the same opportunities.