How the Policy Center Uses Microsimulation Models

Len Burman Urban-Brookings Center and Syracuse University

For presentation at Microsimulation Modelling for Analysis Seville, September 23, 2016

www.taxpolicycenter.org Background on TPC

 Tax Policy Center Founded in 2002  Nonpartisan venture of two respected nonprofit research organizations: The and the  Most of our funding comes from foundations.  We make our research and policy analysis available to the public in clear accessible terms  Our goal is to improve the quality of tax policy discourse in the US

www.taxpolicycenter.org 1 The TPC microsimulation tax model

 We have the same kinds of modeling tools used by government agencies such as CBO, Treasury, and JCT  Many staff have experience working in government agencies

www.taxpolicycenter.org 2 TPC tax model

 Individual calculator – Based on weighted sample of 145,858 tax returns filed in 2006 (Public Use File or PUF) – Inflate/reweight to hit 100 targets for 2011 – Age and extrapolate based on CBO baseline and distributional targets  Add demographic information, nonfilers from CPS  Impute wealth, education, consumption, health, retirement

www.taxpolicycenter.org 3 Other modeled

 Payroll taxes  Assigning burden  Estate tax  taxes  VAT

www.taxpolicycenter.org 4 Uses of model

 Revenue estimating  Distribution  Effective marginal tax rates

See: http://www.taxpolicycenter.org/resources/brief-description-tax-model

www.taxpolicycenter.org 5 The tax system does more than raise revenue

 200 tax expenditures = $1.4 trillion in 2015  Policymakers have lots of new ideas  Modeling them is often a challenge

www.taxpolicycenter.org 6 index of tax expenditure supply: requests to JCT

16,000

14,000

12,000

10,000

8,000

6,000

4,000

2,000

0 1987-88 1989-90 1991-92 1993-94 1995-96 1997-98 1999-00 2001-02 2003-04 2005-06 2007-08 2009-10

Source: Joint Committee on Taxation. Includes “a small number of requests for data other than revenue requests.”

www.taxpolicycenter.org 7 Index of quantity and complexity of tax proposals: length of Administration tax proposal descriptions

www.taxpolicycenter.org 8 Illustration of challenges of tax modeling

 Jeb Bush’s Reform and Growth Act of 2017

www.taxpolicycenter.org 9 Revenue Effect of Bush Individual Income and Proposals, 2016-2026, in $Billions Revenue Provision How to projectChange data forward? Repeal AMT -361 Cut individual income tax rates (max 28%) -2,742 Cut capital income taxes (max rate 20%) -316 Repeal AGI phaseouts (Pease and PEP) -305 Increase standard deduction -815 Increase EITC for childless workers -48 Trim tax expenditures 1,467 relief -527 Social Security tax relief for workers 67 and older -259 Business tax changes -350 Total -4,257 www.taxpolicycenter.org 10 Revenue Effect of Bush Individual Income and Payroll Tax Proposals, 2016-2026, in $Billions Revenue Provision Change Repeal AMT -361

Cut individual income tax rates (max 28%) Easy -2,742 Cut capital income taxes (max rate 20%) -316 if no behavioral Repeal AGI phaseouts (Pease and PEP) -305 response Increase standard deduction -815 Increase EITC for childless workers -48 Trim tax expenditures 1,467 Marriage penalty relief -527 Social Security tax relief for workers 67 and older -259 Business tax changes -350 Total -4,257 www.taxpolicycenter.org 11 Revenue Effect of Bush Individual Income and Payroll Tax Proposals, 2016-2026, in $Billions Revenue Provision Change Repeal AMT -361 Cut individual income tax rates (max 28%) -2,742 Cut capital income taxes (max rate 20%) -316

Repeal AGI phaseouts (Pease and PEP) Model behavioral -305 Increase standard deduction response—short- and -815 long-term Increase EITC for childless workers -48 Trim tax expenditures 1,467 Marriage penalty relief -527 Social Security tax relief for workers 67 and older -259 Business tax changes -350 Total -4,257 www.taxpolicycenter.org 12 Revenue Effect of Bush Individual Income and Payroll Tax Proposals, 2016-2026, in $Billions Revenue Provision Change Repeal AMT -361 Cut individual income tax rates (max 28%) -2,742 Cut capital income taxes (max rate 20%) -316 Repeal AGI phaseouts (Pease and PEP) -305 Increase standard deductionStraightforward to model the sometimes -815 complex decisions for individuals Increase EITC for childless workers -48 Trim tax expenditures 1,467 Marriage penalty relief -527 Social Security tax relief for workers 67 and older -259 Business tax changes But behavioral response to complex incentives is -350 uncertain Total -4,257 www.taxpolicycenter.org 13 Revenue Effect of Bush Individual Income and Payroll Tax Proposals, 2016-2026, in $Billions Revenue Provision Change Repeal AMT -361 Cut individual income tax rates (max 28%) -2,742 Cut capital income taxes (max rate 20%) -316 Repeal AGI phaseouts (Pease and PEP) -305 Increase standard deduction -815 Increase EITC for childless workers -48 Trim tax expenditures Head and spouse earnings not broken 1,467 out on PUF (impute from CPS) Marriage penalty relief -527 Social Security tax relief for workers 67 and older -259 Business tax changes -350 Total -4,257 www.taxpolicycenter.org 14 Revenue Effect of Bush Individual Income and Payroll Tax Proposals, 2016-2026, in $Billions Revenue Provision Change Repeal AMT -361 Cut individual income tax rates (max 28%) -2,742 Cut capital income taxes (max rate 20%) -316 Repeal AGI phaseouts (Pease and PEP) -305 Increase standard deduction -815 Increase EITC for childless workers -48

Trim tax expenditures Age not on PUF (Assign based on IRS 1,467 Marriage penalty relief tabs) -527 Social Security tax relief for workers 67 and older -259 Business tax changes -350 Total -4,257 www.taxpolicycenter.org 15 Special challenges in modeling corporate income taxes

 No PUF  Corporations can be very complex - Many lines of business, spread across many different taxing jurisdictions  Tax info reported to regulators often differs from actual tax liability  We are planning to build a corporate tax model based on COMPUSTAT data and detailed industry tabulations produced by IRS

www.taxpolicycenter.org 16 Revenue Effect of Bush Corporate and Estate Tax Proposals, 2016-2026, in $Billions Revenue Provision Change 20% corporate rate, repeal AMT -1,653 Expensing plus repeal interest deduction -813

Territorial system These provisions would be -159 8.75% tax on foreign accrued profits relatively easy to model 129 with a corporate tax model Trim corporate tax expenditures 245 (Although accurately Permanent R&E credit -81 modeling behavioral Total for corporate income tax response still a challenge.) -2,331 Repeal estate tax + carryover basis -185

www.taxpolicycenter.org 17 Revenue Effect of Bush Corporate and Estate Tax Proposals, 2016-2026, in $Billions Revenue Provision Change 20% corporate rate, repeal AMT -1,653 Expensing plus repeal interest deduction -813 Territorial system Deferred income of -159 multinationals very hard to 8.75% tax on foreign accrued profits model, even with good data 129 Trim corporate tax expenditures 245 Behavioral responses even Permanent R&E credit harder -81 Total for corporate income tax -2,331 Repeal estate tax + carryover basis -185

www.taxpolicycenter.org 18 Revenue Effect of Bush Corporate and Estate Tax Proposals, 2016-2026, in $Billions Revenue Provision Change 20% corporate rate, repeal AMT. -1,653 Expensing plus repeal interest deduction -813 Territorial system -159 8.75% tax on foreign accrued profits 129 Trim corporate tax expenditures 245 No public estate tax return Permanent R&E credit data, but can approximate -81 by killing off people Total for corporate income tax probabilistically in -2,331 Repeal estate tax + carryover basis PUF. -185 Info on carryover basis for one year on 8939 (not public) and SCF.

www.taxpolicycenter.org 19 The changing form of large U.S. businesses

www.taxpolicycenter.org Source: Cooper, et. al (2015) 20 Pass-throughs are very complex

Source: Cooper, et. al (2015) www.taxpolicycenter.org 21 Change in After-Tax Income Under Bush Proposal By expanded cash income percentile, 2017 How to measure 14% income?

12%

10%

8%

6%

4%

2%

0% Lowest Second Middle Fourth 80–90 90–95 95–99 Top 1 Top 0.1 quintile quintile quintile quintile percent percent

Source: Urban-Brookings Tax Policy Center Microsimulation Model (version 0515-3). www.taxpolicycenter.org 22 Measuring economic status

 Income reported on tax returns is incomplete - Missing income excluded from tax (e.g., fringe benefits) - Missing most transfers (e.g., public assistance) - Capital gains reported on “realization basis” - Business income may overstate or understate actual income - Failure to index capital income and expense

www.taxpolicycenter.org 23 Annual income as measure of economic status

 Income varies over the life cycle  Transitory shocks can cause annual income to deviate substantially from “permanent income”  Lifetime income would be a better measure, but requires panel data and way to project future income, which depends on factors not reported on tax returns  Consumption may be better alternative (permanent income hypothesis), but consumption data very noisy  Auerbach, Kotlikoff, Koehler propose alternative method based on SCF data

www.taxpolicycenter.org 24 Challenge in measuring economic incidence

 Corporate income tax  Payroll tax  Economic development tax incentives (claimed by high-income taxpayers, but aimed at providing benefits for low-income households)  Tax-exempt bonds (surprisingly complex incidence)

www.taxpolicycenter.org 25 Other model outputs

 Marginal tax rates on labor, capital income  METR on business investment - based on model computations of average individual tax rates on dividends and capital gains and additional calculations (Marron and Rosenberg, 2015)  Other information relevant to peculiarities of American tax system

www.taxpolicycenter.org 26 Macro modeling issues

 Short-term vs. long-term  Response depends importantly on unspecified items - Central bank response - How and when will deficits be closed?  Highly uncertain parameters and model structure can have large effects on macro feedback (even sometimes the sign)  TPC uses Keynesian model for short-term, OLG model for long-term macro feedbacks - Solow growth model in development

www.taxpolicycenter.org 27 Example: JCT analysis of Dave Camp’s plan

www.taxpolicycenter.org Source: JCT, JCX-22-14 28 Accessing administrative datasets

 IRS makes data available to some researchers subject to very tight restrictions  Some researchers collaborate with analysts inside government agencies that have access to data  PUF – increasingly restricted  Proposal to expand access to researchers outside of government

www.taxpolicycenter.org 29 Much data available to government analysts

 Detailed geographic information  From SSA - Age - Head and spouse earnings - Gender - Earnings histories  Information returns - Retirement accounts - Health insurance - Executive compensation www.taxpolicycenter.org 30 Big data sets

 SOCA  Panel data  Universe of income tax returns filed - Prevalence of e-filing makes these data much more suitable for research than in previous years

www.taxpolicycenter.org 31 Making administrative data available to more researchers

1. Modern computer power makes it possible to create very good synthetic data sets at relatively low cost - Anonymized records designed to match the statistical properties of source data - Synthetic data would have same record structure as underlying administrative data - Different datasets could be designed for different research purposes  E.g., panel data, files representative at state level, files with more detailed asset information

www.taxpolicycenter.org 32 Use and limitations of synthetic files

 Could be used for certain kinds of simple statistical analysis: e.g., OLS and tabulations  More complex models—e.g., bunching models to measure ETI—wouldn’t work unless specifically designed for that purpose  Could use synthetic files to debug statistical programs and then submit the program to the IRS for processing  Fees would cover the cost of dedicated staff that would check for disclosure issues before releasing output www.taxpolicycenter.org 33