 |
|
|
Chairman Calls for Better Federal Budget Estimates
Thursday, May 9, 2002
Statement of Hon. Christopher Cox
Before the House Committee on Rules
Subcommittee on Legislative and Budget Process
Hearing on Assessing the Accuracy of Federal Budget Estimating
Introduction
The Congressional Budget Office (CBO) was created in 1974 to help Congress make wise fiscal policy. Together with the Joint Committee on Taxation (JCT), this significant portion of the Congress’s permanent staff estimates the revenue effects of proposed tax changes. CBO and JCT are failing to provide Congress with useful information. In the past year alone, CBO has revised its long-term estimate of the federal surplus by $4 trillion. JCT calculated the revenue impact of last year’s Economic Growth and Tax Relief Reconciliation Act without even considering its effect on economic growth.[1]
Tax bill scoring is performed by JCT, using CBO economic forecasts as an input. JCT estimators do consider behavioral issues such as tax-avoidance, evasion, and after-tax rates of return in the real world. But they assume that any such effects always cancel each other out. The basic flaw—that is the very cornerstone of JCT’s tax scoring system—is the irrebutable presumption that every tax change will have presumably ZERO effect on the United States economy. No effect on economic growth. No effect on total employment. No effect on total savings and investment. The fatal assumption that the tax code has no effect on the economy produces revenue estimates that are not merely useless, but which actually impede sound tax policy.
Over history, one is hard-pressed to find instances where CBO and Joint Tax predictions have been in the same time zone as actual results. If these Congressional agencies are to provide useful information, they must began by providing accurate information. The debate should not be about static vs. dynamic scoring. It should be about accurate estimates vs. worthless figures that everyone admits are wrong.
If we cannot achieve more accurate scoring, Congress should consider stripping the budget process of this useless appendage.
A History of Error—CBO
A glance at the record demonstrates how the combined errors of JCT tax estimating and CBO economic forecasting have led to completely unreliable predictions of budget deficits and surpluses:
· In 1989, CBO projected a 1990 deficit of $140 billion. The real number was $221.4 billion—an error of 58%[2] in one year.
· In 1990, CBO estimated a 1991 deficit of $138 billion. The actual number was $269.4 billion—an error of 95% in a single year.
· In March 1992, halfway through the fiscal year, CBO made the switch to exaggerated pessimism, and predicted a $368 billion deficit in 1992—an overshoot of $78 billion.
· In January 1993, CBO estimated a deficit of more than $300 billion for 1993—off by more than $50 billion.
· CBO’s same-year deficit predictions for 1994 and 1995 were CBO’s best since I was elected to Congress in 1988—only wrong by a total of $34 billion.
· In mid-1996, CBO predicted a $150 billion deficit for 1996—an overestimate of nearly 50% in the same year.
· In mid-1997, CBO predicted a deficit of $124 billion for 1997—an overestimate of more than 500%.
· In 1998, CBO’s one-year forecast was off by $123 billion—a 98% error.
· In 1999, CBO predicted a surplus of $142 billion for 2000—off by $96 billion, or 68%.
· CBO’s same-year surplus prediction for 2001 was off by $70 billion—or 55%.
The average annual error since I have been in Congress is more than 80%.
Long-term forecasts have been even more erratic. CBO’s estimate for the current fiscal year has ranged from a $579 billion deficit (1993), to a $188 billion deficit (1997), to a $176 billion surplus (2001), now back to a $100 billion deficit (2002). As you can see from this chart, if consistency is the hobgoblin of little minds, then we are blessed with expansive intelligence indeed at CBO.
The most important number for JCT’s scoring of tax bills is the CBO prediction of real GDP growth. CBO recently published an overview of their forecasting record, including a review of their real output growth predictions from 1982-1999. It revealed that their mean error was approximately 22%.[3] CBO’s mean absolute error was approximately 31%. This means that in any given year, CBO’s prediction differed from the actual by an average of 1.0% of real GDP, or around $100 billion.
To demonstrate the utility—or lack thereof—of CBO’s predictions, I have carried out a simple experiment—one I urge you to repeat if you are skeptical. I have substituted Yahtzee dice for CBO’s complex computer simulation. I rolled one die three times for each year, and took the average to produce my “estimate” for real output growth in that year. The dice produced an average growth rate prediction of 3.43%. Compared to the real value of 3.22%, this gives a mean error of 6.5%—a significant improvement over CBO’s effort. The dice had a mean absolute error of 28%—also better than CBO. Perhaps we should turn over responsibility for Congressional budget forecasting to Parker Brothers.
The point of this exercise is not to insult CBO’s economists. In many cases, the errors were the result of factors they could not possibly have foreseen. Their forecasting record is virtually identical to that of the Blue Chip Economic Indicators and other private sector forecasters. The point is that we are spending $30 million a year to reproduce the same Blue Chip numbers we can look up for free on the Wall Street Journal online. The point is that we take too seriously numbers that are not—and perhaps cannot be—more accurate than a roll of the dice.
A History of Error—JCT
Faulty economic and budget forecasts make it hard enough for Congress to implement good tax policy. JCT scoring that deliberately ignores the large impact of tax policy on the economy makes it nearly impossible. This mindset was perhaps best expressed by Laura Tyson, former Chairman of President Clinton’s Council of Economic Advisors, who notoriously said, “…there is no relationship between the levels of taxes a nation pays and its economic performance.”
I do not know how many people working at JCT today agree with Ms. Tyson’s sentiment. But they continue to function as if they do—indeed, as if they hold it true as an article of faith. According to JCT’s system, no matter what we do to taxes, there will be absolutely no effect on jobs, inflation, interest rates, or economic growth. These all must remain unchanged in all circumstances. We all know the world isn’t like that. The historical evidence overwhelmingly confirms that growth in the economy leads directly to higher federal revenues. And there are many examples of economic growth accompanying significant reductions in tax rates:
· When Presidents Harding and Coolidge cut the top marginal rate from 73% to 24% in the 1920s, GNP rose by 60%, and income tax receipts rose by 61% over the next eight years.
· President Kennedy cut taxes on savings and investments in 1962, followed by across-the-board rate reductions in 1964. Over the next seven years, the economy grew by 5% per year, and federal revenues jumped 62%.
· During the 1980s, Congress cut the top income tax rate from 70% to 28%, and the size of the American economy doubled between 1980 and 1990. Tax receipts to the federal government doubled, too, growing by 100% between 1980 and 1990.
The textbook case of JCT’s ignoring the economic effects of taxes is the capital gains tax. For years, CBO and JCT have predicted that reducing capital gains rates would cost the Treasury revenue. Each time they have had it exactly, precisely backwards.
· Between 1978 and 1985, when the tax rate fell from 43.8% to 20%, capital gains tax revenues more than doubled—in real terms.
· When the rate was raised to 28% in 1986, CBO and JCT predicted revenues would rise—they actually fell by 34% in a single year, and by 1991, taxable capital gains were only half of what they had been in 1985.
· In 1997, JCT again predicted that cutting the rate to 20% would cost the treasury billions. Instead, capital gains receipts were $47 billion—90%—higher in 1999 than in 1996. Even with the stock market losses of the past two years, capital gains receipts have been higher with the 20% rate than they were with a 28% rate.
Ignoring economic effects does not just misstate the effects of tax cuts; it also consistently misstates the effects of tax increases:
· In 1990, Congress briefly instituted a tax on luxury items such as yachts. Actual collections from the boat tax were only 60% of JCT predictions. The tax had a disastrous effect on boatyards, driving many workers into unemployment, and thus lowering payroll tax revenues.
· The 1990 Bush tax increase resulted in a loss of revenues in 1991, which defenders of tax increases blame on the recession. Even discounting the fact that the tax increase was one of the major causes of the recession, it is interesting to note that tax revenues from the top tax bracket—the one that was raised—fell by 6%, while revenues from the other brackets continued to rise.
· The 1993 Clinton tax hikes—the largest in history—brought in less than a third of the additional revenue that JCT predicted, and depressed real incomes by $25 billion.
These spectacularly wrong predictions have repeatedly caused Congress to enact tax increases that hamper economic growth and prosperity. Efforts at tax relief have consistently been stymied by unrealistic projections of massive revenue losses. Such a dependable record of faulty forecasts begs for a re-examination of JCT scoring practices.
Objections to Considering the Macroeconomic Effect of Tax Policy
Defenders of the status quo have raised several objections to the consideration of the effects of tax policy on the economy as a whole. CBO Director Dan Crippen concisely stated these objections this morning in his testimony before the House Rules Committee.
The first objection is that it is impossible to predict future fiscal policy, which will undoubtedly also have some effect on future economic performance. Of course, this is no less true using the current models. For estimating purposes, we should start with CBO’s economic forecast, and ask whether the proposed tax rate change will have an effect—and if so, whether it is positive or negative, and how much. If the particular tax cuts result in lower revenues, they’ll lead to additional borrowing, and the resulting rise in interest rates may well cancel out some of the positive economic effects of the tax cuts themselves. Other tax cuts may stimulate both the economy and revenues. Still others may have no positive economic effect, while also losing the government money. All that is needed is to assess the likely effect of the particular tax law change that is proposed.
A second common objection is that Federal Reserve policy would offset the stimulative effects of tax cuts. While it is certainly true that bad Fed policy could harm the economy, it seems unlikely. This would assume—incorrectly—that the Federal Reserve has a target for real output, and would thus thwart any tax policy impact on output.
Director Crippen suggests that taxpayer perception of tax policy is as important as the policy itself, and is impossible to predict. The historical record shows that taxpayers seem to “perceive” a tax cut as a tax cut, and a tax hike as a tax hike, and behave accordingly.
In the end, Director Crippen will admit that of course economic performance is affected by taxes. But he will claim that, since they don’t know how to estimate these effects, it is somehow more objective to simply set them at zero. What makes zero more “objective” than, say, 1, or –1, is not clear. What is clear is that this assumption leads to the gross simplification of predicting tax hikes will always result in higher revenues, and tax cuts will always result in lower revenues.
Recommendations
While no economic model can be 100% accurate, the abysmal record of JCT and CBO scoring compels us to consider alternative forecasting methods. Virtually any alternative would be an improvement over the current system. It is tempting to dispense with the scoring process altogether. After all, Congress operated without scoring for almost 200 years without driving the nation into a fiscal abyss. When the experts say that cutting capital gains rates in 1978 will cause revenues to go down, but they go up; when the experts say that cutting them in 1981 will lower revenues, but again they go up; when in 1986 they say raising capital gains will raise revenues, but they go down; when the experts say yet again that cutting them in 1997 will lower revenues, but yet again they go up; it is impossible not to wonder, who needs the experts? Director Crippen suggests that it is best to assume the economic effects of taxes are always zero, rather than make an inaccurate estimate. I suggest that we extend that reasoning to all of the inaccurate numbers we are getting from CBO and JCT, and assume that the deficit will always be zero, and that the revenue effects of every tax policy will always be zero.
In reality, I think scoring could be a useful tool, and that we should make every effort to improve the process before abandoning it. The first step to improving scoring accuracy is opening up the process to public scrutiny. Currently, JCT does not publish its scoring models. Members of Congress have been attempting to obtain access to JCT models for at least 15 years, without success. It is intolerable that an arm of Congress—intended to help Congress make policy decisions—should conceal its methods of operation. It is particularly strange at a time when Congress is insisting upon accountability and transparency in corporate accounting. The Rules Committee should use its authority to compel JCT to publish their models and equations in comparable detail to a peer-reviewed academic journal. The disclosed information should be sufficient for a knowledgeable outside expert to fully replicate JCT’s calculations.
Until JCT’s models become public, it is impossible to conduct a constructive debate on the many technical issues involved in achieving more accurate scoring. Whatever your opinion of various alternative modeling proposals, everyone can agree that an open and honest system, enlisting the forces of the private sector and academia, would be an improvement. Such openness will also help guarantee that politically motivated modeling choices will be kept to a minimum. As Justice Louis Brandeis said, “Daylight is the best disinfectant.”
Conclusion
Our nation’s long-term economic health depends on sound tax policy. The scoring of tax legislation is meant to help Congress institute this sound policy. Unfortunately, inaccurate scoring has actually been standing in the way of better tax policy for the past three decades. The economy and the American people have been paying the consequences.
No econometric model is perfect, but the JCT and CBO models are almost perfectly wrong. If we in Congress are to have scoring accurate enough to be useful, the macroeconomic effects of tax policy must be taken into account. The inherent complexity and importance of this undertaking requires that it should not be carried out behind closed doors. Therefore, the Rules Committee should require the full disclosure of the models in use. Once these models are made public, the process of review and improvement can begin in earnest. If after all of our efforts it proves impossible to achieve accurate scoring, we must realize that politically loaded numbers that bear no relation to reality are worse than no numbers at all.
[1] Lindy Paull, Chief of Staff of JCT, testified to the House Policy Committee that the economic growth effects of the Bush Tax Cut were not considered when scoring the measure.
[2] Unless otherwise indicated, percentage figures are the error divided by the estimate.
[3] CBO’s mean estimate of real output growth from 1982-1999 was 2.5% annually, whereas the actual figure was 3.22%. Thus, the error of 0.72% is approximately 22% of the actual figure.
|
[Back]
|
|
|
|
