Data Sources

[Oren-T]

There are two main data sources we're thinking about right now:

Bulk data downloads from USASpending.gov

• This is the data source we're currently using
• We may need to use sub-awards down the line, since they often have valuable information, like here (where the main grant is to Cook County but the sub-grants include a ton of more granular information about how the money was actually used/distributed).

OJP funding data

• This comes from here, and is available as an API (easier for real-time integrations).

Gun violence data

• Currently, we are sourcing this data from American Violence
• This uses the past 5 years, looking at total fatal shootings, fatal shooting rates per capita, and population. So currently, this is 2019-2023.

[Oren-T]

There are some validations that need to be done on the current data:

• [x] Make sure the years seem correct
• [ ] Verify that agency filters should, in fact, be awarding agency, not funding agency

[Oren-T]

Potential data source issues

Undercounting:

1. I discovered a funding_office_name called "COMMUNITY ORIENTED POLICING SERVICE," which seems like something that should be included in what we're doing? However, the program_activities_funding_this_award column is blank in the example I have (award is available here). I found it by searching for "chp" in the description.
   • However, it should be noted that some of the "chp" searches are not returning valid cvi grants. For instance, one Chicago program reads in the description, "outreach to five community police officers funded in 2021 by chp in yaboucoa will precede offering cops free, introductory webinars to humacao and maunabo forces." So, it's referencing a different chp grant.
2. Here's another example of that funding source: award seen here, which is a Baltimore grant from 2020. The description just reads "chp" and it also comes from community oriented policing. Again, though, the program_activities_funding_this_award column is blank.
3. Overall, there are 92 grants (filtering on NEW only) in our time period of analysis which contain the text "chp" as a full word in their description, but where the program is not "cops hiring program." 90 of those have a program match of "other." Of those 90, 50 have the description "CHP," 34 are "FY21 COPS HIRING PROGRAM (CHP)," and four are a description of the CHP. So overall, across all cities, we believe we are missing at least 88 grants with our current counting scheme, from CHP alone.

Overcounting

1. We think it is likely that many JAG grants do not relate to CVI programs. For instance, Washington, DC has received over $18.5MM in three JAG grants towards their "Justice For All" initiative: "effective administration of criminal justice training and technical assistance (tta) program." This does not appear to be CVI related, so should not be counted as such.
2. Greensboro has a $300k JAG grant which was used to "purchase equipment that the departments cannot afford otherwise and are not part of their normal budget purchases. these expenditures are meant to enhance service levels, increase officer safety and efficacy." Again, this seems unlikely to be CVI related.

Cities/States/Sub-Grants

Many state capitols get grants which are then allocated to individual counties or cities within the states. As a result, funding numbers for capital cities are likely over-counted, and for non-capitol cities, we are under-counting.

As a temporary fix, we are filtering the column called business_types_description, to remove any grants which are awarded to "STATE GOVERNMENT"s. However, long-term, we seek to incorporate sub-grants as a more accurate way to process these data.

For instance, this VOCA grant to Chicago has 26 sub-awards which specify different cities around the state where the funds were sent. However, there are some small layers of added complexity to consider.

One goal for the data is to present cities with a number of the remaining available funds they have for CVI programs. This becomes difficult to impossible if we are looking at the sub-grant level, since some of the money won't yet have been awarded in sub-grants.

The action date for sub-grants should be greater than or equal to the primary grant date, but this is another thing that needs to be validated.

Once this is all accounted for, however, we should be able to merge together primary and sub-grant data to better understand proper numbers of funding at each city level, ideally resolving much of the over- and under-counting currently happening.

[Oren-T]

Places to correct for over- and under-counting:

• We see two types of cdfa titles related to research: "EDUCATION INNOVATION AND RESEARCH (FORMERLY INVESTING IN INNOVATION (I3) FUND)" and "NATIONAL INSTITUTE OF JUSTICE RESEARCH, EVALUATION, AND DEVELOPMENT PROJECT GRANTS." Spot checking and GPT validation suggests that grants in these two categories are overwhelmingly research only, and so could be excluded. (However, it should be noted that many are RCTs so may relate to actual program evaluation. Do we want to filter those as well?)

[Oren-T]

American Violence data corrections:

There are three "cities" listed in the American Violence data set which I am excluding due to them not having city-level alignment:

'Louisville/Jefferson County, Kentucky',
'Nashville-Davidson, Tennessee',
'Urban Honolulu CDP, Hawaii'

[Oren-T]

Baton rouge has some grants we exclude:

grants_to_exclude = [ 'https://www.usaspending.gov/award/ASST_NON_B-21-DF-22-0001_8620/', 'https://www.usaspending.gov/award/ASST_NON_B-18-DP-22-0001_8620/', 'https://www.usaspending.gov/award/ASST_NON_B-21-DZ-22-0001_8620/', 'https://www.usaspending.gov/award/ASST_NON_B-22-DF-22-0001_8620/', ]