Yesterday, the Sunlight Foundation launched Clearspending, an assessment of the federal government's spending reporting. As Sunlight's Ellen Miller explains, the investigation found widespread problems with data quality:

The data inaccuracies we uncovered account for 70 percent of the total $1.9 trillion in government spending data reported in that year. Some of the numbers are too big, some are too small and some are missing completely, while other spending data entries don’t have the detail that’s required or were reported months later than the law demands. (emphasis added)

Specifically, Clearspending looked at three metrics:

• Consistency: Does the spending reported on USASpending.gov match the numbers in the Catalog of Federal Domestic Assistance (CFDA)? The CFDA contains estimates of annual program spending, while USASpending displays spending actually reported.
• Completeness: Was information entered in all required data fields? The Federal Funding Accountability and Transparency Act (FFATA) requires information on the recipient, funding agency, amount, location the work was performed, and so on.
• Timeliness: Was the spending reported on time? FFATA requires that spending be reported within 30 days.

The study found widespread errors in each category. For instance, student loan spending was reported at $5 trillion -- more than the entire federal budget (and approximately 8% of the world's GDP). The average delay between obligating spending and reporting it was 55 days, nearly double the required 30 day limit imposed by FFATA.

There are some limitations to the study, particularly with regard to the consistency metric:

Since CFDA program obligations are annual estimates, there are very few programs whose reported CFDA obligation total is exactly equal to the aggregate obligations in USASpending.gov. As a result, the presence of over- or underreporting cannot be assumed to represent an error. ... We are unable to offer an accuracy measurement due to the difficulty of tracing an obligation’s origin to the budget.

Nevertheless, Clearspending begins to paint a picture of agency compliance with FFATA, and the results are not pretty.

The administration has recognized spending data quality as a concern. OMB issued memos on the topic in February and April of this year. Each agency has a Senior Accountable Official for spending data quality, and each agency must establish a data quality plan (see e.g. EPA's plan).

Despite these efforts, serious data quality issues persist. USASpending can be a powerful tool for transparency, but only if the underlying data are complete, accurate, and timely. We need answers as to why there are such widespread errors in the data — and how policies and practice will be changed to improve it.