kbmorales/ppp

An R Package that downloads Paycheck Protection Program loan data and reads it in as a tidy data set

PPP

The goal of PPP is to download and assemble a unified, cleaned data set of Paycheck Protection Program loans issued in 2020.

The data is too large to share on GitHub, but this package will allow you to recreate the data locally.

The work for this package grew out of a project with DataKind’s DC chapter and the National Press Foundation.

Installation

You can install the development version from GitHub with:

# install.packages("devtools")
devtools::install_github("kbmorales/PPP")

Assembling PPP data

# Note: this function will download over 600MB of data to your local machine,
# and read a large dataframe into memory!
ppp_data = ppp_assemble()

Data Sources

• Small Business Administration PPP Loan Data:
  • Latest release
  • Original release
• NAICS code dictionaries US Census NAICS Files

PPP Data Dictionary

Though there are two versions of the PPP data released, both will have the same final structure as detailed below. The main structual difference between these versions is the removal of a JobsRetained variable, and the addition of JobsReported. Whether or not this is a simple semantic change is unknown at this point.

The final PPP dataset output by ppp_assemble() will contain the following columns:

variable	original	notes
LoanRange	Yes	no values for loan amounts under 150K
BusinessName	Yes	no values for loan amounts under 150K
Address	Yes	no values for loan amounts under 150K
City	Yes
State	Yes
Zip	Yes
NAICSCode	Yes
BusinessType	Yes
RaceEthnicity	Yes	89.3% “Unanswered”
Gender	Yes	77.7% “Unanswered”
Veteran	Yes	84.7% “Unanswered”
NonProfit	Yes
JobsRetained	Yes	only in 2020-07-06 release
JobsReported	Yes	only in 2020-08-08 release
DateApproved	Yes
Lender	Yes
CD	Yes
LoanAmount	Yes	no values for loan amounts over 150K
source_file	No	file name for where data was pulled
version	No	release date of PPP data
LoanRange_Unified	No	Loan ranges regardless of amount
JobsRetained_Grouped	No	brackets for # of ‘jobs retained’
JobsReported_Grouped	No	brackets for # of ‘jobs reported’
LoanRangeMin	No	minimum possible loan value
LoanRangeMax	No	maximum possible loan value
LoanRangeMid	No	middle estimated loan value
naics_lvl_1	No	Most general NAICS industry class
naics_lvl_2	No	Second level NAICS industry class
naics_lvl_3	No	Third level NAICS industry class
naics_lvl_4	No	Forth level NAICS industry class
naics_lvl_5	No	Most specific NAICS industry class
NAICS_version	No	version of NAICS where code was found
NAICS_valid	No	was NAICS code matched?

ppp's Issues

fallback for URL download fail

1. naics files bundled with package?
2. check if url is still good
3. offer option to provide your own URL
4. offer option to use a local file

build code to handle a local file if user prefers

edit COLLECT script

Set up tests

We need tests for the R package set up with testthat

Ones in particular that would be super useful:

1. Check status of download URLs for raw data files to ensure they still return a 200 status code before attempting to download
2. Checking number of rows before and after joins to ensure join duplication does not occur
3. Ensure that coercion of variables in ppp_clean does not lose information (e.g., turning LoanAmount into a double doesn't turn some useful character data into a NA

ppp_assemble(version = 1) --- crash!

Runs.
But no error, just no data.
Nor are the data directories set up.

(Have not traced through code yet).

I mostly use vanilla terminal, nvim and without any interface

qualified names (package::FUN())

Not a biggie, but R-pkg book, Ch 6 and Ch 13.6.1 recommends:

All of our calls to tidyverse functions have now been qualified with the name of the specific package that actually provides the function, e.g. dplyr::mutate(). There are other ways to access functions in another package, explained in chapter 13, but this is our recommended default. It is also our strong recommendation that no one depend on the tidyverse meta-package in a package10. Instead, it is better to identify the specific package(s) you actually use. In this case, our package only uses dplyr.