This document describes Subsidyscope’s methodology for measuring spending on federal programs that contain subsidies and the subsidies they provide. This methodology is organized in three parts. First, it describes the important difference between measuring spending on programs that contain subsidies and measuring actual subsidies. Second, it describes, in four steps, how Subsidyscope identifies programs that contain subsidies. Third, it describes how Subsidyscope collects spending data on these programs and, where available, data on subsidies.
A. Subsidies and Programs That Contain Subsidies
In order to measure subsidies, one must identify those programs that produce them. Unfortunately, there is no single definition of the term “subsidy” and people disagree about which federal programs do or do not produce subsidies. In order to inform our thinking, Subsidyscope uses, as a starting point, the Government Accountability Office’s definition of "subsidy":
Generally, a payment or benefit made by the federal government where the benefit exceeds the cost to the beneficiary. Subsidies are designed to support the conduct of an economic enterprise or activity …. They may also refer to (1) provisions in the tax laws for certain tax expenditures and (2) the provision of loans, goods, and services to the public at prices lower than market value. These include interest subsidies.
Even this definition is quite general, and as straightforward as it may sound, interpreting it is not a simple task. Subsidies are not often discernible line items in the federal budget; rather, the subsidy is a portion of the amount spent on programs that provide subsidies. The amount spent on programs containing subsidies is, in turn, a subset of all government spending (see the figure below). Within total government spending, Subsidyscope first identifies spending on government programs that contain subsidies.
Subsidies are a subset of programs that contain subsidies; these programs are, in turn, a subset of all government programs
(for illustration only, not based on actual data)
Defining what government programs provide a subsidy is a matter of judgment. Determining what activities “support the conduct of an economic enterprise or activity” as the GAO definition suggests can be difficult. For instance, some government activities provide specific benefits to specific people, not to an “enterprise or activity.” Governments may also indirectly support an economic activity while in pursuit of a broad or diffuse policy goal, such as public safety.
Not all spending on programs that provide a subsidy translates into a subsidy. The subsidy is the net monetary benefit a recipient receives from a program.1 In some cases this may be significantly less than what the federal program actually spends. For example, a policy holder of federal flood insurance may receive $10,000 from the government to cover a claim. While the federal government has spent $10,000, the net benefit to the policy holder would be less if the policy holder paid insurance premiums to the government before filing the claim.
Unless a government agency or some other entity calculates what portion of spending on subsidy programs is a subsidy, the exact amount of subsidy may be unknown. Subsidyscope presents government data on the programs that contain subsidies, and data on specific subsidies, when available. When subsidy estimates do not exist, we present the amount of spending on the programs containing subsidies, providing an upper bound for the estimate of the actual subsidy. However, government data quality issues explored further here lead us to believe that the estimates we provide for spending on programs that contain a subsidy are lower bound estimates for the total spending.
It is important to note this project does not include all government subsidies. First, Subsidyscope only looks at federal programs. Subsidies provided by state or local governments are not considered unless they originate from federal funding. Second, subsidies provided by regulations or tariffs are not included in the scope of the project. Finally, Subsidyscope is not attempting to measure negative subsidies (e.g., excise taxes).
B. Identification of Programs That Contain Subsidies
In order to estimate spending on programs that provide subsidies and the subsidies they produce, Subsidyscope must identify what federal programs contain subsidies. To do this, Subsidyscope follows a four-step process.
- Categories of Subsidy Programs.
The federal government can deliver subsidies six different ways. These are listed below. Subsidyscope starts our process by looking for government sources of subsidy information in the first four categories on this list.
- grants (a direct expenditure);
- contracts (a direct expenditure);
- tax expenditures;
- risk transfers (loans, loan guarantees, and insurance);
- regulatory subsidies; and
- tariffs/excise taxes.
Direct expenditures include direct transfers of money (e.g., cash grants) or goods and services (e.g., donation of government surplus).2 Government contracts are used to provide services for the government, either on behalf of the government or for the government directly. Under a contract, a subsidy generally occurs when the government pays more than fair market value for a good or service. Subsidyscope presumes that competed contracts—contracts that are subject to an open bidding process—generally do not have a subsidy component, even though the bidding process may include certain preferences, such as for minority-owned businesses. Therefore, we focus our data presentation on non-competed contracts. Tax expenditures provide tax relief to certain parties by allowing special exclusions, exemptions, deductions, special credits, preferential rates or deferrals of liability, before taxes are calculated. Risk transfers convey financial risk from recipients to the federal government through insurance contracts, loans, loan guarantees and similar instruments.3
Importantly, the last two categories of subsidies, regulatory subsidies and tariffs/excise taxes, are not measured by Subsidyscope. Regulations affect market prices and have the effect of subsidizing firms or individuals that are not regulated or can help existing firms that are regulated by discouraging new competition. Likewise, tariffs and excise taxes – or fees charged on certain goods – are a negative subsidy, discouraging the purchase of goods subject to the tariffs or tax. While Subsidyscope recognizes that they can have a significant impact on economic behavior, these categories are excluded from our scope to keep the project manageable.
- Sector Selection
Second, Subsidyscope selects a particular economic sector on which to focus. In order to define a sector in a way that lends itself to data collection and aggregation, we begin by using established federal budget functions to define sectors. (An exception to this, the nonprofit sector, is described in the paragraph below.) Using budget functions makes it easier to define the scope of a sector and identify federal programs in that sector. For instance, the transportation sector is budget function 400 and the energy sector is budget function 270. However, there are many exceptions as some activities within a sector cut across multiple budget functions. We use a variety of means to identify these cross-cutting programs and apportion their costs to appropriate sectors.
Subsidyscope examined one sector that is not defined by a budget function: the nonprofit sector. Subsidyscope is an initiative of the Pew Charitable Trusts and is implemented in conjunction with the Sunlight Foundation. Like other nonprofits, the Pew Charitable Trusts and Sunlight benefit from federal subsidies. However, since there is no budget function for nonprofits, these subsidies might otherwise be spread out among other sectors or not covered at all. Subsidyscope treats nonprofits as a separate sector. Thus, Subsidyscope provides the same detail and summary subsidy information for nonprofits as it provides for other sectors. The parameters used to define this sector are described on the nonprofit sector summary page.
- Identification of Federal Programs in a Sector
Third, the Subsidyscope team identifies federal programs in a selected sector (e.g., transportation) by using the first four categories, created in the first step above, and various government data sources. This is done as follows:
- Direct expenditure programs are identified from the Catalog of Federal Domestic Assistance (CFDA) first using budget functions, and second, using the CFDA functional index4 to further identify cross-cutting programs that may pertain to a specific sector. Subsidyscope manually sorts the programs in the functional index to determine whether to include those programs in the sector.
- Contracts for a sector are identified from the Federal Procurement Data System (FPDS), and FPDS-NG (Next Generation), which is a part of USAspending.gov. We use North American Industry Classification System (NAICS) codes to identify the appropriate economic sector, or Product Service Codes, if NAICS codes are not available.
- Tax expenditure programs are identified from the Analytical Perspectives volume of the President’s Budget.
- Risk transfer programs are identified from the Analytical Perspectives volume and the Federal Credit Supplement of the President’s Budget, and other agency-specific sources.
Using this information, Subsidyscope compiles a list of federal programs (including non-competed contracts but excluding regulations, tariffs and excise taxes) that are directed at a particular sector of the economy.
- Identification of Programs That Contain a Subsidy
Finally, having identified the federal programs within a specific sector that may fall into one of the four subsidy categories, Subsidyscope identifies the subset of these programs that are likely to contain a subsidy.
Subsidyscope uses a decision tree (shown above) to identify those federal activities that are likely to contain a subsidy. The decision tree categorizes activities of the federal government based on both the type of activity and the characteristics that define a subsidy. For instance, some federal activities, such as the collection of tariffs, fall outside the purview of the project and, therefore, are not categorized as activities that include a subsidy component for the purpose of Subsidyscope analyses. Activities within the scope of the project are sorted through a cascading set of options which help identify the type of activity (e.g., contract, risk transfer or tax expenditure) and whether or not that activity meets two important characteristics of a subsidy as defined by GAO: (1) does it result in a net benefit to the beneficiary and (2) does is it support the conduct of an economic activity?
It is important to note that the decision tree does not eliminate the need for judgment. Some judgments are built into the tree. For instance, the decision tree assumes all tax expenditures provided to entities other than individuals (e.g., corporations) contain a subsidy component. It also assumes that all federal contracts provided at ‘market value’, but where that market is constrained (e.g., not open to full competition) are more likely to contain a subsidy component. Other judgments must be made in this sorting process. Most notably in many cases a decision must be made whether or not a federal activity supports “the conduct of an economic enterprise or activity.” In some cases this is not a difficult question. For instance, a federal tax credit for buying residential solar panels clearly supports undertaking an economic activity (the production of solar panels). But other cases are not as clear. Does a federal grant to a police department to generally promote automobile safety, including the option to use some of the money to purchase child car seats for low income families, constitute supporting “the conduct of an economic enterprise or activity”? In this case Subsidyscope decided it did not, because the purchase of car seats seemed a small and optional part of the overall program which did not otherwise seem to support any other economic activity. But others could disagree with this answer.
Subsidyscope’s data set does not preclude others from using our decision tree or their own judgment to come up with different conclusions. In order to provide wide usability, and to account for different interpretations, we present all government direct expenditure data within a sector in our search tools, except in the area of fully competed contracts [see more in Sec. C below]. This allows users to, for instance, include any programs in their database search and count them as subsidies or omit other programs that we have excluded. In this way, users may make their own judgments, or change the scope of their research, and conduct their own analyses.
C. Estimating Spending on Subsidies and Programs That Contain Subsidies
Having identified federal programs that are likely to contain a subsidy in a sector, Subsidyscope obtains estimates of the amount spent on these programs and the amount of subsidy delivered to the recipients, where available. This is done by mining different government databases and documents. The sources depend on the category of the program as defined in sub-section B.1. above, and requires a different process of determining subsidies for each type of federal spending.
Direct expenditures on grant programs are often the easiest on which to collect data. However, subsidies delivered through direct expenditures are often the most difficult to identify, given that not all of the money spent through direct expenditures is a subsidy, and measuring the subsidy requires economic data that may or may not be available (such as a market price in the absence of government spending).
To that end, Subsidyscope presents the grant data obtained through USAspending.gov in a way that allows users to sort the data across various categories. The data are already classified by the government according to assistance type (grants, direct payment, loan guarantee, etc.), as well as by recipient type (state, county, nonprofit, etc.). It is important to note that most spending data presented are obligations.
Data on government contracts are pulled from USAspending.gov, which houses the Federal Procurement Data System (FPDS). Subsidyscope does not attempt to determine which contracts have a subsidy component or measure what that component may be. We do presume, however, that competed contracts—contracts that are subject to an open bidding process—generally do not have a subsidy component, even though the bidding process may include certain preferences. Thus, we concentrate on those contracts that are non-competed and make this information available to the public. This does not mean non-competed contracts contain a significant subsidy, only that it is more likely to be the case.
Our definition of “non-competed contracts" includes all six categories of non-competed/ constrained competition in FPDS (all categories except “A Full & Open Competition,” “F Competed under Simplified Acquisition Threshold” and “CDO Competitive Delivery Order”).
Contracts data are presented separately from grants and should be viewed with additional caution for at least three reasons. First, many contracts, even those where competition is limited, are awarded at close to fair market value such that the actual subsidy provided may be low or zero. Second, Subsidyscope is unable to use budget functions to organize these data into sectors. As mentioned above, we are using North American Industry Classification System (NAICS) codes to identify the appropriate economic sector, or Product Service Codes, if NAICS codes are not available. Because contracts cannot be sorted by budget function, Subsidyscope is reluctant to combine contract data with expenditure data. Finally, Subsidyscope has found the quality of contract data in USAspending.gov to be very poor. For instance, despite there being a requirement that contracting officers identify NAICS codes for each contract, Subsidyscope found significant variation in reporting across years. For example, in fiscal year 2000, approximately 88 percent of all contract records are missing NAICS codes while in fiscal year 2010, only 3 percent of records are missing NAICS codes. This lowers our confidence in the use of NAICS codes going back in time for the contract data. In part, this is why the use of Product Service Codes (PSC) became a supplement to the NAICS codes.
For tax expenditures, Subsidyscope uses data from the U.S. Department of the Treasury’s estimates presented in the Analytical Perspectives of the President’s Budget. As previously noted, the subsidy provided through a tax expenditure may not be 100 percent of the estimated revenue loss of the tax expenditure. For example, certain tax expenditures place compliance costs on the recipient which may reduce the value of the tax expenditure. However, compliance costs are not calculated, therefore Subsidyscope presents the entire amount of a tax expenditure as a tax subsidy.
Estimated credit subsidies provided through risk transfers come from a number of sources, including the Analytical Perspectives volume and the Federal Credit Supplement of the President’s Budget, and other agency-specific sources. Under the Federal Credit Reform Act, the federal government requires the subsidy costs of loans and loan guarantees to be estimated and presented in the budget. Estimates of subsidies provided through insurance programs are not required and thus harder to find. For several noteworthy examples, specifically in the Agriculture and Housing sectors, Subsidyscope provides a brief description of insurance programs.
Data on government contracts are pulled from USAspending.gov, which houses the Federal Procurement Data System. These data are presented separately from the other categories and should be viewed with additional caution for at least three reasons. First, many contracts, even those where competition is limited, are awarded at close to fair market value such that the actual subsidy provided may be low or zero. Second, Subsidyscope is unable to use budget functions to organize these data into sectors. As mentioned above, we are using North American Industry Classification System (NAICS) codes to identify the appropriate economic sector, or Product Service Codes, if NAICS codes are not available. Because contracts cannot be sorted by budget function, Subsidyscope is reluctant to combine contract data with expenditure data.
Of course, all of Subsidyscope’s estimates of subsidy spending and subsidies rely on the quality of data provided by the government databases. Overall we believe the estimates of spending on grant programs and contracts to be lower bounds due to missing data in federal data sets; however, it is important to reiterate that these spending estimates represent upper bounds for the subsidy estimates contained within those grant programs or contracts. To the extent these data are found to be of poor quality or omit important information, Subsidyscope will attempt to obtain more accurate information directly from agencies or other sources of raw data. For graphic presentations of data on our site where we know the omissions are significant, such as subsidy estimates for direct expenditures, we will not supply an estimate but insert a question mark or a statement such as "subsidy unknown."
- Some argue that a government subsidy occurs when the government fails to correct problems that result in certain goods or services being under- or over-priced. For instance, they argue that a government’s failure to adequately reduce greenhouse gas emissions results in coal being under priced; resulting in a subsidy for coal companies. Subsidyscope does not take this approach. In practice, it would be very difficult to identify and measure unaddressed market failures. Deciding who should get how much from the government to make the world more equitable is a central function of government. Subsidyscope makes no judgment regarding whether a particular program or policy is equitable
- Subsidyscope’s definition of direct expenditures includes all programs designated federal assistance types "A" through "D" and types "H" through "M" in the Catalog of Federal Domestic Assistance.
- Subsidyscope’s definition of risk transfers includes all programs designated federal assistance types "E", "F" or "G" in the Catalog of Federal Domestic Assistance.
- The National Defense sector is unique in that it has no functional index.