Transparency troubles at Recovery.gov

When the American Recovery and Reconstruction Act passed back in February, its backers promised an unparalleled level of disclosure about where the money is going. While I applaud the sentiment, the release of stimulus data in the last month has been anything but smooth.

It’s a classic case of over-promising and under-delivering.

First, it was the job numbers. Now comes the phantom Congressional districts.

Since much of the direct stimulus aid goes to state capitals, any attempt to analyze stimulus spending by Congressional district alone would skew the figures. In the chart below, about half of the stimulus awards in Oklahoma — $1.3 billion — is going to the 5th Congressional District, which includes Oklahoma City.

After our story came out today, one reader e-mailed and said recipients could have entered their state House or Senate districts by mistake instead of the Congressional districts. For me, more troubling than the so-called “fake” Congressional districts is the fact that a good chunk of the data don’t include any Congressional district at all (the “Null” field above).

Still, in the wake of all the negative press about the Congressional districts, the officials behind Recovery.gov said they have updated the data on the site that erroneously placed stimulus awards in nonexistent districts.

If a recipient reported an incorrect or invalid congressional district, the code “ZZ” appears in the “Congressional District” field as a placeholder. The recipient will change  the report with the correct congressional district during the next reporting period, beginning January 1, 2010.

Also, the federal watchdogs in charge of stimulus spending were grilled on Capitol Hill this morning. In a related report, the Government Accountability Office had some interesting things to say about how stimulus transparency has fared so far. What I found interesting is its investigators started doing error analysis on the same day the data was released to the general public.

GAO performed an initial set of basic analyses on the final recipient report data that first became available at www.recovery.gov on October 30, 2009; reviewed documents; interviewed relevant state and federal officials; and conducted fieldwork in selected states, focusing on a sample of highway and education projects.

While much of the blame for erroneous or missing data is being placed at the recipients who filled out the Web forms on FederalReporting.gov, many are asking if there shouldn’t be some type of validation for the data before it’s released to the public.

The guidance released back in the summer by the Office of Management and Budget lays out who is responsible for data quality in stimulus reporting.

Data quality is an important responsibility of key stakeholders identified in the Recovery Act. Prime recipients, as owners of the data submitted, have the principal responsibility for the quality of the information submitted. Sub-recipients delegated to report on behalf of prime recipients share in this responsibility. Agencies funding Recovery Act projects and activities provide a layer of oversight that augments recipient data quality. Oversight authorities including the OMB, the Recovery Board, and Federal agency Inspectors General also have roles to play in data quality. The general public and non-governmental entities interested in “good government” can help with data quality, as well, by highlighting problems for correction.

To be fair, more than 100,000 reports on stimulus spending have been filed so far. Some data entry errors or miscategorizations were bound to happen when you have that many people filling in Web forms. Here’s what the GAO said it found:

Our review also identified a number of cases in which other anomalies suggest a need for review: discrepancies between award amounts and the amounts reported as received, implausible amounts, or misidentification of awarding agencies. While these occurred in a relatively small number of cases, they indicate the need for further data quality efforts.

Officials hope that future stimulus reports will contain fewer errors as recipients become more comfortable filling out the forms and the requirements are refined. In the meantime, Recovery.gov could make an easy fix by allowing users of the site to flag data that plainly looks wrong. It could function much like the “star ratings” systems on other sites, where users could “grade” their view of the data accuracy in a report.

For more on Recovery.gov data quality, check out the Sunlight Foundation and ProPublica. The White House also published a lengthy retort to critics earlier this week on its blog.

–Paul

UPDATE, 6:10 p.m., 11/19/09: From the prepared remarks of the Earl Devaney, chairman of the Recovery Accountability and Transparency Board:

These mistakes do not surprise me, however, and in a serendipitious way, they are not unequivocally bad. In reality, this data should serve in the long run as evidence of what transparency can achieve.

In the past, this data would have been scrubbed from top to bottom before its release, and the agencies would never have released the information until it was perfect. You — and the American public — are now seeing what agencies have seen, internally, in the past. And what we are all seeing, at least following this first reporting period, is not particularly pretty.

This raw-form, unsanitized data may cause embarrassment for some agencies and recipients, but my expectation is that any embarrassment suffered will encourage self-correcting behavior and lead to more accurate reporting in the future.

Written by Paul Monies

Follow @pmonies

Thank you for joining our conversation on Data Watch. We encourage your discussion but ask that you stay within the bounds of our commenting and posting policy.

Powered by Wordpress | WP Premium theme by WP Remix
Copyright 2007. Data Watch. All rights reserved

• Home
• RSS Feed
• About
• NewsOK | Right to Know