了解CBO经济预测中的生产率增长

The NABE Foundation’s23rd Annual Economic Measurement Seminar Aaron BetzMacroeconomic Analysis Division For information about the seminar, seewww.nabe.com/ems2026. Purpose of CBO’s Economic Forecast The forecast is used primarily as an input to CBO’s 10-year federal budgetprojections and analyses of legislative proposals. It is a current-law forecast: It reflects the assumption that current laws governingtaxes and spending generally will not change but that any policy changesscheduled under current law will take effect as planned. For example, one tax provision in recently enacted legislation allows for thededuction of up to $40,400 of state and local taxes in 2026. Under current law, thatamount will fall to $10,000 after 2029. CBO’s current-law forecast incorporates theeconomic responses to that scheduled decline beginning in 2030. CBO’s Approach to Forecasting CBO’s approach involves projecting: ▪Potential output(that is, maximum sustainable gross domestic product, orGDP), by using a Solow-type growth model, and ▪Actual output,by using alarge-scalemacroeconometricmodel. The estimate of potential output is based mainly on estimates of: ▪Thepotential labor force, ▪Theflow of services from the capital stock, and ▪Potential total factor productivity (TFP)in the nonfarm business sector. The ratio of real (inflation-adjusted) potential GDP to the potential labor force isknown aspotential labor force productivity. Average Annual Growth of Real Potential GDP and Its Components Key Estimates in CBO’s Projection of Potential GDP,February 2026 Percent Total Factor Productivity in the Nonfarm Business SectorSince 2000 Index, 2000 = 1.0 Federal Debt Held by the Public in CBO’s Baseline andUnder Scenarios With Faster or Slower TFP Growth Percentage of GDP Potential Improvements to CBO’s MethodofForecastingTFP CBO’s Current Approach to Forecasting TFP inthe Nonfarm Business Sector Estimate potential, or trend, TFP over history.The agency uses the latestavailable data to estimate past trends. ▪Potential TFP is estimated using geometric growth trends over the peak-to-peak business cycles from 1948 to 2007. ▪Those trends are estimated in a piecewise linear (or “jointed stick”) regression. ▪The estimates of potential TFP are alsoused to determine the most recenttrend in potential TFP growth beginning with the start of the last full businesscycle (the fourth quarter of 2007). Project near-term potential TFP growth.The most recent trendgrowth rate ofpotential TFP is assumed to converge linearly over several years to theweightedaverage of potentialTFP growth over the past 25years. Project long-term potential TFP growth.After the convergenceperiod,potential TFP grows at a rate equal to the 25-year weighted average. Reasons for Revisiting CBO’s TFP Forecasting Model The main limitation of the agency’scurrent method of estimating and projectingpotential TFPisthat it is slow to recognize and incorporate changes in underlyingtrends. Changes may take severalyears to be fully reflected in the agency’s forecast. That issue has become more pronouncedin thepostpandemicperiod during whichmeasured TFP growth has exceeded its average rate of growth from the fourthquarter of 2007 to the fourth quarter of 2019. The agency is evaluating various other methods of estimating historical potential TFPand projecting future TFP growth, including unobserved-components methods,Bayesian vector autoregressions, and newer approaches, such as those developedby Ulrich Müller and Mark Watson. CBO is also exploring incorporating a labor quality measure into its estimate of laborsupply, which would improve estimates of historical TFP by treating changes in thequality of the workforce as increases in labor input rather than as productivity gains. Criteria for Evaluating TFP Forecasting Methods The primary criterion for evaluating TFP forecasting models is their out-of-sampleforecast performance, as measured by absolute and mean-squared errors. CBO also places some weight on the stability of models’ estimates of historicalpotential TFP. Those estimates should not be overly sensitive to recentobservations. The agency’s objective is to improve the accuracy of its forecasts of TFP whileavoiding excessive sensitivity to recent observations in its estimates of historicalpotential TFP. How Artificial Intelligence MightAffect Productivity How Does Automation Affect Labor Productivity? The shared framework of Acemoglu (2024), Aghion and Bunel (2024), and Arnon(2025) describes how the automation of tasks by AI affects labor forceproductivity: Where: 1−𝛼is the labor share of income, ϴis an occupation-specific measure ofexposure to generativeAIautomationbased on automatable tasks, 𝜌is a measure ofadoptionrepresenting the percentage of exposed tasks thatare economically attractive to automate, and 𝜇is a measure of thecost savingsresulting from automation. Calculating the Effect of Task Automation on Productivity This task