Read a CBO cost estimate yourself before you believe the headline

Here is a number you will see on cable news within an hour of any major bill passing a committee: "The bill will cost two trillion dollars." Or: "The bill is fully paid for." Or, my favorite, "The bill will cut the deficit," said by a person who has never once voted to cut anything. Each of those sentences usually traces back to the same twelve-page document, and almost nobody quoting it has read past the first table.

You can. That is the good news. The document is public, it is written in a plain if dry style, and a patient citizen with an hour can learn to read it better than most of the people paid to describe it. The document is the cost estimate from the Congressional Budget Office, and this post is a manual for reading one. The steps are numbered and plain. I will keep the commentary light and save the argument for the end, where I will tell you why I think this skill is among the most useful a conservative citizen can have.

Who the CBO is, and what it is not

Congress created the Congressional Budget Office in the Congressional Budget and Impoundment Control Act of 1974. Its job is to give Congress budget numbers that do not come from the party leadership or the White House. Its director is appointed jointly by the Speaker of the House and the President pro tempore of the Senate, after considering recommendations from the two Budget Committees, and its stated mission is objective, nonpartisan analysis. It scores what the bill text says, under a set of rules, and reports a number.

Keep two things in your head. First, the CBO is not an oracle. It estimates the future, and its estimates have been wrong in both directions, as every forecast is. Second, the CBO is not a villain. When it tells a sponsor that a bill costs more than advertised, the sponsor will call it biased; when it tells the opposition the same thing, they will call it a hero. Judge the agency by whether you understand its method, not by whose ox it gores.

The CBO produces cost estimates for most bills that a committee approves, and it publishes them for free on its website, cbo.gov. A separate body, the Joint Committee on Taxation, which is a committee of Congress with its own professional staff, estimates the revenue effects of tax provisions. You will need both for any bill that touches taxes. More on that in step six.

Step 1: Find the estimate on the CBO's website

Go to cbo.gov and use the search box. Type the bill number, for example the House or Senate bill number as printed in Congress.gov, or the short title. Look for the item labeled a cost estimate, not a blog post or a testimony. The estimates are filed by bill and by date, and a bill that changes in committee can have more than one estimate: one for the bill as introduced, one for the version as ordered reported, and sometimes one for the version that passed a chamber. Make sure you are reading the estimate for the text you care about. The date and the version appear at the top of the first page, usually in a line saying the bill was "as ordered reported by" a named committee on a named date.

If you cannot find an estimate, that is itself information. The bill may be too early in the process, or the sponsor may be moving it without the committee review that triggers a score. A bill that has not been scored has no official cost, and any figure you hear for it is somebody's guess.

Step 2: Read the summary table first

The first page or two of an estimate usually carries a short summary and a table titled something like "Estimated Budgetary Effects." This is the table that matters. It lists, year by year for the next five and ten years, the change in the federal deficit that the bill would cause, with a column or two of totals. A negative number in the table means a reduction in the deficit or an increase in revenue. A positive number means more spending or less revenue.

Read the rows before the totals. A typical table will have rows for "Direct Spending" and "Revenues," and a bottom line for the "Net Increase or Decrease in the Deficit." Direct spending, which the budget calls mandatory spending, is money that flows under existing law once the bill passes, without a later appropriation vote. Revenue is what the Treasury collects. Look at the bottom line, then look at how the years behave. Does the line start at zero and climb? Does it fall in the early years and rise in the late years? The shape tells you more than the total does.

A note on the "On-budget" and "Off-budget" lines you will sometimes see. Social Security's trust fund transactions are classed as off-budget by law, so an estimate may show a bill's effect on the on-budget deficit separately from its effect on Social Security. When a sponsor says "no effect on the deficit," check whether they mean on-budget only.

Step 3: Tell budget authority from outlays

Washington uses two words that sound alike and mean different things. Budget authority is the legal permission Congress gives an agency to commit the government to spend, by signing contracts or making grants. Outlays are the actual cash that leaves the Treasury. A bill may grant billions in budget authority that is spent over many years. When the CBO says a program "costs" a number in a given year, it usually means outlays in that year.

Why does this matter to you? Because a program can look small in the first year's outlays and large in total budget authority, and because a multi-year commitment, such as a contract or a loan guarantee, can be reported in a way that hides the later bills. When you see a line labeled "Authorization" or "Estimated Authorization," check whether the amount is subject to appropriation. Authorizations of appropriations are permission to spend later if Congress votes the money. They are not a commitment. Direct spending is the opposite: it is the commitment.

Step 4: Find the ten-year window, and the gimmicks it invites

Nearly every estimate covers a ten-year window, a budget convention and, for the reconciliation procedure in the Senate, a rule of thumb that effects beyond the window count against a bill. The window is a convenient yardstick. It is also a target, and legislators who want a bill to look cheap will design it to fit the window.

Here are the three most common ways to game ten years, so that you know what to look for.

1. Start late. A new benefit that does not begin until year four or five costs nothing in the early years. The ten-year total shows only six or seven years of costs, and the full annual price appears after the window closes.
2. Stop early. A tax cut or spending program that expires after five years shows five years of cost, even if everyone in the room expects Congress to extend it. The estimate follows the text of the law, and the text says it ends.
3. Shift the timing. Moving a payment from the last day of one fiscal year to the first day of the next shifts the cost from one year to another without changing the total. A bill can look cheaper in one window by pushing costs into the next.

The window is not a lie. A forecast of the next ten years is fair and useful. It becomes a gimmick when the design of the bill is driven by the window rather than by the policy.

Step 5: Read the sunsets and the baseline

Look in the estimate for a sentence beginning "The bill would" and then for words such as "expire," "terminate," "sunset" or "temporary." Then ask what the estimate assumed would happen when the provision ended. The answer is built into the CBO's rules. The CBO measures a bill against its baseline, a projection of what federal spending and revenue would be if current law stayed in place. Under the baseline's rules, a tax provision that expires under current law is assumed to be gone. Expiring mandatory spending programs are treated differently: many are assumed to continue, so the two sides of the ledger do not behave alike.

That produces a surprising result. If a tax cut is set to expire and a bill extends it, the CBO scores the extension as a cost, because the baseline assumes the tax goes back up. The sponsor says "this is not a tax cut, it is keeping taxes where they are." The estimate says "this reduces revenue by so many billions compared with the law as it stands." Both are describing something real. They are using different yardsticks. Read both versions and decide which fits your question.

Spending is scored against the baseline too. For a discussion of how baselines can turn a smaller increase into a reported cut, see baseline budgeting, where a smaller increase gets reported as a cut.

Sunsets are not only a trick of the sponsor. Senate rules in the reconciliation process, under what is called the Byrd rule, bar provisions that add to the deficit beyond the budget window. That is how the 2001 tax cuts came to carry an expiration date at the end of 2010. A provision that sunsets may be a deliberate choice or a procedural necessity. The estimate will not tell you which. The bill's history will.

Step 6: Check for dynamic analysis, and compare with the Joint Committee on Taxation

Most estimates are conventional. They hold the overall size of the economy fixed and count the direct effects of the bill on spending and revenue, with some behavioral responses folded in. A tax cut is scored as lost revenue, with an allowance for how people change their behavior, but not for economic growth that might result.

For very large bills, an additional dynamic analysis may be published. It asks how the bill would change the economy's output, employment and interest rates, and how those changes would feed back into revenue. Dynamic analysis is not an automatic bonus. It carries more uncertainty than the conventional estimate, and its results depend on assumptions about how people respond. Treat it as a second opinion, to be read alongside the first, never instead of it. When a politician from either party cites "growth effects," ask whether there is a published dynamic estimate, who prepared it, and how large the effect was relative to the total.

For anything that changes taxes, the Joint Committee on Taxation produces the official revenue table. It appears as a table of "Estimated Budget Effects" and is published on the JCT's website, jct.gov, in a numbered series of documents. Compare its totals to the CBO's. The two are designed to match, since the CBO uses the JCT's revenue numbers in its own estimate, and a mismatch is a flag that you are looking at different versions of the bill.

Step 7: Check what the estimate says about mandates, interest and uncertainty

The last pages of an estimate hold the details most readers skip. Look for three items.

• Mandates. Under the Unfunded Mandates Reform Act of 1995, the CBO states whether a bill contains intergovernmental or private-sector mandates and whether their cost exceeds a statutory threshold. If your city or state will have to pay for something Congress ordered, this section says how much. It is a good place to start if you care about the states and their budgets.
• Interest. Some estimates give a separate line for the added interest cost of financing the deficit the bill creates. If a bill adds $100 billion in borrowing, the Treasury pays interest on it for years. The line may be small in the first years and large later.
• Uncertainty. An estimate usually says what could make it wrong: a different interest rate, different participation in a program, a different court ruling. When the document describes the estimate as "uncertain" and tells you why, take the hint.

A worked example with a made-up bill

Let me show you the steps on a hypothetical bill. This is invented for teaching and corresponds to no real legislation. Call it the Family Widget Act. It does two things. It creates a widget subsidy for families, and it extends a widget tax credit that is set to expire.

Suppose the estimate's table reads like this, in billions of dollars:

The headline writer sees $290 billion and calls it a modest sum. Apply the steps.

Step 2: the bottom line is $290 billion over ten years. The shape is the story: $30 billion in the first three years, $60 billion in years four and five, and $200 billion in years six through ten. The cost is climbing.

Step 4: look at the timing. The subsidy has no cost in years one through three. Suppose the bill's text says the subsidy starts in year four. In years six through ten the subsidy runs at a pace of $200 billion over five years, or $40 billion a year. That is the full annual price, and it has five years of data in the window and then continues. If the program continues at $40 billion a year for the next ten years after the window, it costs another $400 billion. The ten-year total of $240 billion understates the long-run cost by more than half. Check the arithmetic: five years at $40 billion is $200 billion, and the two earlier years at $20 billion each is $40 billion, so $240 billion in all. The window shows one dollar of cost for every roughly two the program will incur over the following decade.

Step 5: the credit extension runs only through year five. If the estimate assumes it expires then, the cost stops at $50 billion. But suppose the sponsors plan to extend it again, as Congress often does with expiring credits. Ten years of the same credit would cost about $50 billion for the first five years and, at the same pace, about another $50 billion for the next five years, for a total of $100 billion, double the reported number. The sunset made the credit look half as expensive as it would be if it were kept.

Step 6: for the credit, which is a tax provision, find the JCT table and confirm the $50 billion matches. If it does not, find out why.

So what does the honest summary say? Not "$290 billion." It says: the bill costs $290 billion in the window as written, would cost roughly $290 billion plus about $400 billion more for the subsidy and $50 billion more for the credit if both were continued beyond the window, and its cost grows over time. That is a different story, and you found it with arithmetic and a table.

Why a conservative should care

Let me argue the point for a moment, because it is the reason I wrote this. The framers of the Constitution gave Congress the power of the purse in Article I, section 9, which says that "No Money shall be drawn from the Treasury, but in Consequence of Appropriations made by Law; and a regular Statement and Account of the Receipts and Expenditures of all public Money shall be published from time to time." A regular statement and account. That is the Founders' demand that the public be able to see what the government spends. A cost estimate is the modern version of that statement, and the reading of it is a citizen's part of the bargain.

Here is where I part company with the people who treat the CBO as an enemy. The score is the best available defense against an old trick, the claim that a bill is free. Large spending bills often arrive with a promise that they pay for themselves, and the table behind such a promise is where you check it. The debt that results is not an abstraction; see Seven myths about the national debt that Washington finds convenient and track the total on the national debt clock. The remedy is not to attack the referee. It is to read the scoreboard.

And yes, it cuts against some conservative claims as well. When a tax cut is described as paying for itself, the JCT table and the dynamic analysis will tell you how much of it, if any, comes back; the honest ledger is at where tax cuts pay for themselves and where they honestly do not. A reader who wants to win arguments for limited government has to win them on the numbers, and that means not flinching when the numbers disagree. Related reading on the structural side is a balanced budget amendment, the case for it and the conservative doubts, and on why the borrowing limit is not the brake people think it is, the debt ceiling began in 1917 to make borrowing easier, not harder.

The case in brief

1. Every major bill gets a CBO cost estimate, free on cbo.gov, and the Joint Committee on Taxation supplies the revenue table for tax provisions.
2. Read the summary table first, and read the shape of the yearly numbers before the total.
3. Know budget authority from outlays, and authorizations from direct spending.
4. The ten-year window invites late starts, early sunsets and timing shifts, so check what happens in year eleven.
5. The baseline is current law, so extending an expiring provision scores as a cost, and dynamic analysis is a second opinion, not a replacement.

What you can do: pick one bill your representative has voted on, find its CBO estimate at cbo.gov, and send your office a one-sentence question about the year-six-through-ten line.