oldfart
Older than dirt
Recently Wendy Edelberg, Assistant Director for CBO’s Macroeconomic Analysis Division, made a presentation at the Conference of Business Economists on “Dynamic Scoring” on May 14, 2015. Bunches of people who already know all about it and don’t want to be bothered with the details can safely ignore this post. People seriously interested in how CBO makes it estimates should be interested.
The 2016 Budget Resolution altered the procedures for scoring bills by instructing the Congressional Budget Office (CBO) and Joint Committee on Taxation (JCT) to incorporate to the greatest extent practicable the budgetary effects of changes in macroeconomic variables resulting from legislation (other than appropriations acts) that either has a gross budgetary effect of 0.25% of GDP in any year over the next ten years (currently about $45 billion) or is designated by the Chairman of one of the Budget Committees. The estimates are to also include the budgetary effects including macroeconomic effects for the subsequent 20 year period.
Prior to this most legislation was subject only to “static” scoring which ignored macroeconomic effects of budgetary measures. Dynamic scoring has been used for the President’s Budget, the annual long term budget outlook, analysis of fiscal policy scenarios, and one specific piece of legislation (S. 744, the Immigration Act).
What dynamic scoring means is that the budget estimates will include “feedback” mechanisms. In addition to the direct effects of fiscal policy, the CBO will estimate the effects stemming from changes in behavior of consumers, businesses, and investors as they react to those changes. For example, if Congress were to approve a program to spend $1 trillion over ten years to rebuild highways, bridges, and tunnels, there would be a direct effect from the spending of $1 trillion. But such an increase in demand would also increase employment, and the new employees would spend much of those increased wages and businesses would spend more on raw materials and equipment. These effects are measured by a “demand multiplier”.
In addition to these fiscal effects, some measures, such as S. 744, would have significant impact on the labor supply. So would changes in the Social Security retirement age or benefit formula, rules for SSA disability benefits, government and military pensions, and so forth. Energy, environmental, and agricultural measures also would have impact through substitution effects. A “cap-and-trade” regime on carbon emissions would encourage CO2 sequestration to replace fracking, power plants to retool to use gas rather than coal, and nuclear, wind, and solar energy to substitute for fossil fuels. So almost any kind of legislation can have macroeconomic effects, even if it is “revenue-neutral”.
CBO analyzes each proposal in at least two time frames. The short term analysis focuses on changes in direct demand for goods and services, taking into account the gap between production and maximum sustainable (“potential”) GDP. When the economy is very close to capacity, increases in demand in the short term are likely to increase price levels for the goods and services involved, while when there are wider gaps the effect is mainly in increased production as underutilized capacity is put back into production. Personally I am skeptical of this kind of narrative by itself. Different industry structures can create inflation in some markets while not others as a result of the same project. Suppose we implement a major highway program. In terms of the labor market, the main effect could be increased employment and production with little inflationary impact. But at the same time, it could cause prices to rise in “bottleneck” resources such as heavy equipment production. In the long term fiscal policies affect production mainly through altering the savings rate, federal investment, long term business investment, and household decisions regarding work and education.
For the short term analysis, changes in government spending focus on the “demand multiplier”. This measures the increase in spending secondary to the initial spending. CBO makes a case that this indirect spending is highly sensitive to monetary policy. They estimate that if there is minimal monetary policy response (no tightening of credit by the Fed) that over four quarters the demand multiplier will range from 0.5 to 2.5 with a central estimate of 1.5. So over a year a $10 billion spending program can be expected to result in $25 billion of total demand. When monetary policy is more restrictive, when the Fed tightens money, the four quarter demand multiplier will range from 0.4 to 1.9 with a central estimate of 1.2.
For the longer term, CBO uses two models of potential output to estimate fiscal effects over the longer haul; a “Solow”-type growth model” and a life-cycle growth model, which are complementary. These link economic growth to the size of the labor force, the amount of capital, the growth of technology and the efficiency of organizational structure. In a Solow-type model, there is a “wage elasticity” of after-tax income, i.e. how much the labor supply will grow due to a 1% increase in earnings. CBO estimates this figure to be about 0.19 (composed of a substitution elasticity of .24 and an income elasticity of -.05). So if total wages increase by 1%, the labor force should increase 0.19%. In a Solow-type model, as income (not just earned income) increases, so does the savings. This component is very sensitive to distributions of income as the savings rate varies by income group.
The life cycle model adds increased savings due to the household sector protecting itself against future drops in income or large expenses. With a rapidly growing older population, this effect is pronounced.
So is “dynamic” scoring better than static scoring? In theory the answer has to be yes. For small projects, it really makes little difference, but for large changes over 10 to 20 years, the effects can be dramatically different. Of course, the further out we go, the less reliable the predictions become. And with the passage of time technology changes and the behavioral assumptions underlying economic models change. I don’t think that anything over ten years is worth much (except for certain demographic projections, the cohort of 20-year olds will be the cohort of 40-year olds in another 20 years!)With dynamic scoring, we are introducing economic modeling and parameters which are difficult to measure and increase the anticipated degree of error. So what’s the answer? My instinct tells me that properly done, dynamic modeling is a good first approximation of what we are trying to get at. I anticipate that compared to static modeling, dynamic modeling will give us answers in the same direction with a slightly higher amplitude. Any other result would tend to call into question the design of the particular dynamic modeling itself.
The 2016 Budget Resolution altered the procedures for scoring bills by instructing the Congressional Budget Office (CBO) and Joint Committee on Taxation (JCT) to incorporate to the greatest extent practicable the budgetary effects of changes in macroeconomic variables resulting from legislation (other than appropriations acts) that either has a gross budgetary effect of 0.25% of GDP in any year over the next ten years (currently about $45 billion) or is designated by the Chairman of one of the Budget Committees. The estimates are to also include the budgetary effects including macroeconomic effects for the subsequent 20 year period.
Prior to this most legislation was subject only to “static” scoring which ignored macroeconomic effects of budgetary measures. Dynamic scoring has been used for the President’s Budget, the annual long term budget outlook, analysis of fiscal policy scenarios, and one specific piece of legislation (S. 744, the Immigration Act).
What dynamic scoring means is that the budget estimates will include “feedback” mechanisms. In addition to the direct effects of fiscal policy, the CBO will estimate the effects stemming from changes in behavior of consumers, businesses, and investors as they react to those changes. For example, if Congress were to approve a program to spend $1 trillion over ten years to rebuild highways, bridges, and tunnels, there would be a direct effect from the spending of $1 trillion. But such an increase in demand would also increase employment, and the new employees would spend much of those increased wages and businesses would spend more on raw materials and equipment. These effects are measured by a “demand multiplier”.
In addition to these fiscal effects, some measures, such as S. 744, would have significant impact on the labor supply. So would changes in the Social Security retirement age or benefit formula, rules for SSA disability benefits, government and military pensions, and so forth. Energy, environmental, and agricultural measures also would have impact through substitution effects. A “cap-and-trade” regime on carbon emissions would encourage CO2 sequestration to replace fracking, power plants to retool to use gas rather than coal, and nuclear, wind, and solar energy to substitute for fossil fuels. So almost any kind of legislation can have macroeconomic effects, even if it is “revenue-neutral”.
CBO analyzes each proposal in at least two time frames. The short term analysis focuses on changes in direct demand for goods and services, taking into account the gap between production and maximum sustainable (“potential”) GDP. When the economy is very close to capacity, increases in demand in the short term are likely to increase price levels for the goods and services involved, while when there are wider gaps the effect is mainly in increased production as underutilized capacity is put back into production. Personally I am skeptical of this kind of narrative by itself. Different industry structures can create inflation in some markets while not others as a result of the same project. Suppose we implement a major highway program. In terms of the labor market, the main effect could be increased employment and production with little inflationary impact. But at the same time, it could cause prices to rise in “bottleneck” resources such as heavy equipment production. In the long term fiscal policies affect production mainly through altering the savings rate, federal investment, long term business investment, and household decisions regarding work and education.
For the short term analysis, changes in government spending focus on the “demand multiplier”. This measures the increase in spending secondary to the initial spending. CBO makes a case that this indirect spending is highly sensitive to monetary policy. They estimate that if there is minimal monetary policy response (no tightening of credit by the Fed) that over four quarters the demand multiplier will range from 0.5 to 2.5 with a central estimate of 1.5. So over a year a $10 billion spending program can be expected to result in $25 billion of total demand. When monetary policy is more restrictive, when the Fed tightens money, the four quarter demand multiplier will range from 0.4 to 1.9 with a central estimate of 1.2.
For the longer term, CBO uses two models of potential output to estimate fiscal effects over the longer haul; a “Solow”-type growth model” and a life-cycle growth model, which are complementary. These link economic growth to the size of the labor force, the amount of capital, the growth of technology and the efficiency of organizational structure. In a Solow-type model, there is a “wage elasticity” of after-tax income, i.e. how much the labor supply will grow due to a 1% increase in earnings. CBO estimates this figure to be about 0.19 (composed of a substitution elasticity of .24 and an income elasticity of -.05). So if total wages increase by 1%, the labor force should increase 0.19%. In a Solow-type model, as income (not just earned income) increases, so does the savings. This component is very sensitive to distributions of income as the savings rate varies by income group.
The life cycle model adds increased savings due to the household sector protecting itself against future drops in income or large expenses. With a rapidly growing older population, this effect is pronounced.
So is “dynamic” scoring better than static scoring? In theory the answer has to be yes. For small projects, it really makes little difference, but for large changes over 10 to 20 years, the effects can be dramatically different. Of course, the further out we go, the less reliable the predictions become. And with the passage of time technology changes and the behavioral assumptions underlying economic models change. I don’t think that anything over ten years is worth much (except for certain demographic projections, the cohort of 20-year olds will be the cohort of 40-year olds in another 20 years!)With dynamic scoring, we are introducing economic modeling and parameters which are difficult to measure and increase the anticipated degree of error. So what’s the answer? My instinct tells me that properly done, dynamic modeling is a good first approximation of what we are trying to get at. I anticipate that compared to static modeling, dynamic modeling will give us answers in the same direction with a slightly higher amplitude. Any other result would tend to call into question the design of the particular dynamic modeling itself.