Stimulus Transparency of Recovery.Gov

Written by Bandit | 03 February 2010

What started out as a technical discussion, one IT system vs another, it turned into something that caught our eye:  An interesting numerical excavation.   We leave it to you to determine the political value.

Here's our question going into this:   From a mostly IT perspective, is the current RECOVERY.GOV site/data reliable and consistent?

Yes.  It's possible to draw those conclusions from a visitor-level fly-by.

Putting aside the debate of open source versus proprietary software, recovery.gov has all the whiz bangs to navigate and "drill down" into the stimulus data.   (FYI, here's the problem with whiz bangs:  This is what's known as IT-slight-of-hand.   I swear sometimes I think JAVA was created to take our eye off the ball.   All that mouse-over Web 2.0 cool stuff distracts us from the meat that should be on the bone.)

Step 1 - Get Data

Our first step was to get the raw data, and bless their hearts, it was there.  Located in the FAQ section is the ability to download the entire stimulus data set from February 09 through December 09 into an excel spreadsheet.  First blush reveals 258,273 rows of data that has allocated $285,475,811,090.93 of stimulus monies and some data validity issues.

• Invalid zip codes (not 5 or 9 digits)
• Empty project names
• Empty project descriptions
• Duplicate projects due to typo in project numbers

Yesterdays Millions Are Todays Billions

After looking over stimulus data, it quickly became a mind-numbing experience seeing the amount of dollars being allocated for certain projects.  When you think back in time, it wasn't that long ago that millions was considered a big number in daily conversations and that gave way to billions.  With the introduction to stimulus as daily conversation piece, billions have given way to trillions.  Now, trillions are a daily discussion point with health care, government budgets, and deficits.  How soon before we are discussing quadrillion?  Since our unfunded liabilities of roughly 107 trillion, we are already 1/10 of the way to a quadrillion.   (Most people don't even know what comes after a billion)

I digress, back to stimulus insight.  Since recovery.gov allows you to drill right down to a city and project level, I decided to step back and take a overview perspective of the overall numbers and not focus on the political nature of each project.

Step 2 - Simple Numerical Analysis of the Data

• First, there is still roughly $500+ billion that has not been allocated
• There are 74 categories of recipient states.   These include the 50 US states, 22 non US states, other, and empty(blank)
• Some examples of non US states:  Guam, Somoa, and Virgin Islands
• Using Governors as a guide post:  Of the 74 recipient states, 26 would be considered Rep, 29 would be considered Dem, and 19 were not affiliated
• 68,196 projects awarded of the 258,273 fell into the empty(blank) recipient state category to the tune of $9.2 billion
• Projects allocated outside the US: 979  Value:  $3.2 billion

Further analysis focusing on the 50 states and Washington DC.:

• For those interested in trivia, did you know the Secretary of the Army plays the role of governor for DC?  Since he reports to the President, I allocated DC as a Dem affiliate
• Currently there are 24 Rep and 27 Dem Governors which represents the changes in Va and NJ.  For analysis purposes, I went back to Feb. 09 when the stimulus was written, where the governor break down was 22 Rep and 29 Dem.
• Red States - 73,837 projects for $118.6 billion which is 41.5% of stimulus
• Blue States - 115,261 projects for $154.2 billion which is 54%
• Let's take a look at the impact of California, and propose that CA is actually a Blue state with Governator aside.  CA led all states in projects and monies allocated.
  
  • Red States - 57,605 projects for $86.6 billion which is 30.3% of stimulus
  • Blue States - 131,493 projects for $186.3 billion which is 65.3%

Let's take a look at the Top 10

The Conclusion

The data is interesting, both what you can gather from the front end (the website) and what you can see from the downloaded version.     There are many discussions you could start with this, but we shall stick to our opening question:   Does the system provide accurate transparency of stimulus spending?

To this we say:  Absolutely yes, Kind of and Not at all

In the Yes category you have a website that allows you to see lots of dots sprinkled over a geographical map.   At the US level, the entire map is covered with dots, so maybe they need to get some smaller dots.    When you drill into a city, you can actually use those dots to obtain more information about what money was allocated/spent there.    Should you isolate a dot of interest, you can obtain the primary and all sub-awards for that location.

Kinda of because from the website, you really can't find out how the money was spent.  eg.,  In a particular school-related item for Michigan, you can see that many millions went to a single dot.   However, the sub-award detail can at times, not match the total and/or the detail provided leaves you with no sense of knowledge.    e.g.,  In a K-12 school project there was a sub-award to a local computer vendor.    It was for about $30,000 and included a single line of detail:  A teaching-related computer.     There was no ability to drill into ANY level of detail beyond that type of vague single line variety.    We looked at about a dozen sub projects in this same primary award and found two things of interest:   1) ALL of the sub awards were +- $3K of $30K.    From a gut-level audit instinct, that just seemed odd.   2)  No sub project had any meaningful project info that told how or where the money was spent.

Not at all because in some cases, the data isn't there.    It's missing or mis-entered.   Zip codes in the raw data that connect to 9999, but the zip code search on the front end will only accept 5 digits.    68,000+ funded projects that have no state connected to them.  If they keep you from searching for bad data on the website, isn't there a way to keep people from also entering bad data?    We conclude not.    Is it transparent to have $9 BILLION worth of projects with no real connecting data?

Overall:   Not So Good. From an IT perspective, it would get a D- because if your only mission is to provide transparency to spending, then columns and rows MUST cross-foot.   You can't have lots of bogus data on the backend.    Allowing 9999 as a zip code is a bug that should have been fixed long ago.    (And maybe it was, but then all those records in the raw data should have been updated as well)