Poverty and inequality effects of new unemployment benefits, payroll tax cuts, and universal payments¶
By Max Ghenis and Nate Golden
July 2020
Abstract¶
The expiration of the $600-per-week Federal Pandemic Unemployment Compensation (FPUC) program at the end of July 2020 has spurred discussions for new Covid-19 relief legislation in the United States. Proposals for this bill include extending FPUC throughout 2020, cutting payroll taxes, and unconditional payments. We use data from the Current Population Survey between 2009 and 2018 to estimate the poverty and inequality effects of a hypothetical FPUC and FPUC extension, compared to budget-neutral payroll tax cuts and universal payments, respectively. Our simulations show that a universal payment would have lowered poverty and inequality less than April-to-July FPUC (i.e., the policy that’s expiring). However, universal payments would have reduced poverty and inequality more than August-to-December FPUC extensions; for example, it would have reduced poverty 24 percent more in 2009. These deltas are larger for the poverty rate of children, Black people, and people with disabilities. Payroll tax cuts would have had much smaller effects than either policy, and universal payments that include children are consistently more effective than adult-only payments.
Background¶
The CARES Act established two major new unemployment insurance reforms: Pandemic Unemployment Assistance (PUA), which expands unemployment benefits to more worker categories, and Federal Pandemic Unemployment Compensation (FPUC), which adds a federally-funded $600 per week to all unemployment benefits. While PUA is available for up to 39 weeks through the end of the year, FPUC runs only from April through July.
Given FPUC’s expiration and the Congressional Budget Office’s July 2020 projections that the unemployment rate would remain above 10 percent throughout 2020, federal lawmakers are preparing for a new relief package. Congressional Democrats are seeking to extend FPUC in full through January, as specified in the House-passed HEROES Act. The Trump administration has proposed cutting payroll taxes, and members of both parties have spoken of another round of direct payments, following the Recovery Rebates in the CARES Act.
FPUC’s work disincentives have made it controversial: it replaces over 100 percent of wages for all workers earning below $21 per hour, and depending on the state, for workers earning up to $33 per hour.1 Capping the wage replacement rate was infeasible due to limitations with state unemployment benefit computer systems. Its proponents emphasize income stabilization and the need to reduce work to prevent Covid infections. To the extent that FPUC may be making it harder to find workers, payroll tax cuts can offset that effect.
Distributional analysis has been less prominent in the debates over Covid relief policies. Columbia University’s Center on Poverty and Social Policy estimated that the CARES Act roughly offsets the poverty caused by the pandemic’s reduction in market incomes.
Our analysis estimates FPUC’s poverty effects in the CARES Act period, as well as FPUC’s potential expansion through the rest of the year, compared to budget-neutral payroll tax cuts and universal payments.
Data¶
We use the Current Population Survey March Supplement from 2009 to 2018, obtained from IPUMS. The Supplemental Poverty Measure was introduced in 2009, while 2018 is the latest survey year available as of this writing.
We extract fields used for tax and poverty analysis, as well as the “number of weeks, in single weeks, that the respondent looked for work or was on layoff during the preceding calendar year.” This field allows us to estimate the total FPUC benefits for each respondent. We also extract the disability status (i.e., whether the respondent reports any physical or cognitive difficulty), which the CPS collects only for civilians aged 15+.
All dollar values are scaled to 2020 dollars using FRED’s CPI-U series.
Methodology¶
Imputing FPUC unemployment benefits¶
The CPS only provides the number of weeks each respondent was unemployed, not the specific calendar weeks. Since the specific time periods—April to July to model past policy, and August to December to model proposed policy—are relevant for the simulation, we provide each person $600 per week of unemployment, multiplied by the number of weeks in the program (17 for FPUC, 22 for FPUC extension) divided by 52 total weeks. This is the asymptotic equivalent to imputing specific weeks among the total weeks unemployed.2
Tax simulation¶
We use the open-source Tax-Calculator software for simulating FPUC tax liabilities. This involves grouping people into tax units3 and calculating tax-relevant fields from each tax unit’s aggregate CPS responses. We then calculate each tax unit’s federal tax liability with and without FPUC—applying 2020 tax law for all years—and store the difference as the FPUC tax.
To identify the net-of-tax FPUC benefits for each person,4 we allocate the tax unit’s FPUC tax across individuals in proportion to their FPUC benefits. This leaves us with the net-of-tax FPUC benefits for each person in the sample.
Calculating budget-neutral alternatives¶
For each year, we calculated the total cost of FPUC and expanding FPUC by adding up the costs for each person. This table shows these costs, along with other quantities from the CPS that factor into the budget-neutral parameters, for each year. As the economy improved from 2009 to 2018, and specifically as the unemployment rate fell, the cost of FPUC and expanding FPUC fell by roughly half. For example, enacting FPUC from April to July of 2009 would have cost an estimated $68 billion (total), while it would have cost $32 billion in 2018.
"""
SETUP
"""
import numpy as np
import pandas as pd
import taxcalc as tc
import microdf as mdf
import plotly.express as px
import pandas_datareader as pdr # For CPI series.
person = pd.read_csv('fpuc/data/cps.csv.gz')
# These hold functions.
%run 'fpuc/convert_asec_taxcalc.py'
%run 'fpuc/make_tax_units.py'
# This runs execution, and also includes more functions.
# -i gives it access to the active environment (person and prior functions).
%run -i 'fpuc/run.py'
Matplotlib is building the font cache; this may take a moment.
# Print overall summary
print("All figures in millions.")
(year_summary.rename(columns={'FLPDYR': 'Year'}).set_index(
'Year')[OVERALL_YEARLY_METRICS].rename(
columns=DISPLAY_METRICS) / 1e6).style.format("{:,.0f}")
All figures in millions.
| Cost of FPUC | Cost of expanding FPUC | Population | Adult population | Total FICA | |
|---|---|---|---|---|---|
| Year | |||||
| 2009 | 67,500 | 85,630 | 304 | 230 | 565,285 |
| 2010 | 63,310 | 80,186 | 307 | 232 | 554,978 |
| 2011 | 60,172 | 76,143 | 309 | 235 | 430,923 |
| 2012 | 55,853 | 70,642 | 311 | 237 | 429,108 |
| 2013 | 51,203 | 64,709 | 313 | 239 | 580,207 |
| 2014 | 42,712 | 54,093 | 316 | 242 | 590,995 |
| 2015 | 39,791 | 50,398 | 319 | 245 | 643,590 |
| 2016 | 36,919 | 46,757 | 320 | 246 | 654,402 |
| 2017 | 32,277 | 40,924 | 323 | 249 | 672,941 |
| 2018 | 31,933 | 40,440 | 324 | 251 | 687,864 |
These figures produce the budget-neutral policy parameters for each year by division. For example, to calculate the universal payment associated with FPUC in 2009, we can divide the cost of FPUC in that year ($68 billion) by the population in that year (304 million), and arrive at $222 per person. Similarly, the associated FICA cut would be the total FICA collections ($68 billion) divided by $118 billion, or 12 percent (note this is the cut for FICA through the full year).
This table shows the full set of parameters that match the total cost of both the original FPUC and its expansion (“FPUC2”).
# Print reform parameter summary.
(year_summary.rename(columns={'FLPDYR': 'Year'}).set_index('Year')[
FPUC_YEARLY_METRICS + FPUC2_YEARLY_METRICS].rename(
columns=DISPLAY_METRICS)).style.format("{:,.0f}")
| Universal one-time payment (FPUC) | Adult one-time payment (FPUC) | FICA % cut (FPUC) | Universal one-time payment (FPUC2) | Adult one-time payment (FPUC2) | FICA % cut (FPUC2) | |
|---|---|---|---|---|---|---|
| Year | ||||||
| 2009 | 222 | 294 | 12 | 281 | 373 | 15 |
| 2010 | 207 | 273 | 11 | 262 | 345 | 14 |
| 2011 | 195 | 256 | 14 | 247 | 324 | 18 |
| 2012 | 180 | 236 | 13 | 227 | 298 | 16 |
| 2013 | 163 | 214 | 9 | 206 | 270 | 11 |
| 2014 | 135 | 176 | 7 | 171 | 223 | 9 |
| 2015 | 125 | 163 | 6 | 158 | 206 | 8 |
| 2016 | 115 | 150 | 6 | 146 | 190 | 7 |
| 2017 | 100 | 130 | 5 | 127 | 164 | 6 |
| 2018 | 98 | 127 | 5 | 125 | 161 | 6 |
Poverty estimation¶
SPM poverty rates are calculated by grouping individuals into SPM units, which are sub-household groups that the Bureau of Labor Statistics identifies as sharing resources. Each SPM unit then has its own poverty threshold—based on composition and housing situation—and resources—based on income, taxes, mandatory spending categories, and transfers. SPM units are determined to classified as poor if their poverty threshold exceeds their resources.
To calculate this, we group each person into their SPM unit (each year) and aggregate resources under each of the nine conditions:
Baseline
FPUC (April through July) and budget-neutral alternatives:
Payroll tax cut budget
Equal per-person payment
Equal per-adult payment
FPUC extension (August through December) and budget-neutral alternatives:
Payroll tax cut budget
Equal per-person payment
Equal per-adult payment
These SPM-unit resource totals facilitate calculation of the poverty gap (the total amount of money it would take to end poverty, if perfectly targeted). Calculating the poverty rate requires merging resource back to the person level on SPM unit ID, as the poverty rate is calculated as the share of people (not SPM units) in poverty.
We compute inequality as the per-person Gini index, based on their SPM unit’s per-capita resources.
Results¶
We begin by showing the effects of FPUC from April to July, compared to budget-neutral payroll tax cuts and universal payments, and then show the effects of expanding FPUC from August to December.
Policies are evaluated against the following metrics:
Poverty rate: The share of the population in SPM units whose poverty threshold exceeds their resources.
Poverty gap: The total dollar amount required to raise each SPM unit to its poverty threshold (this is a measure of poverty depth).
Gini index: Measure of inequality, calculated from the person-level distribution of SPM unit total resources.
FPUC thus far¶
FPUC and universal payments would have been more progressive than payroll tax cuts, and in most outcomes and years, FPUC would have been slightly more progressive than universal payments, which would have been about 20 percent more progressive than adult-only payments.
Poverty rate¶
In most of the last decade, FPUC would have reduced poverty than the alternatives. However, in 2009, when economic conditions most closely resembled today’s, universal transfers were nearly equivalently effective. The antipoverty effect of all policies shrinks over time, as unemployment has fallen and therefore a hypothetical FPUC’s budgetary impact has shrunk.
In all years, universal transfers and FPUC outperformed payments to adults only, and especially outperformed the payroll tax cut.
# Round for the hover text.
line_graph(pov_rates2[(pov_rates2.age_group == 'All') &
(pov_rates2.race == 'All') &
(pov_rates2.disability_filter == False)],
group='fpuc', y='pov_pc',
yaxis_title='Change in SPM poverty rate',
title='Poverty reduction by reform')
Poverty gap¶
The poverty gap is an alternative poverty measure which quantifies the total amount we’d have to distribute to end poverty. Since filling the poverty gap requires giving more to people who are farther below the poverty line, this measure goes beyond the poverty rate in also measuring the severity of poverty.
The poverty gap trend resembles the poverty rate trend.
# Round for the hover text.
line_graph(pov_gap_ineq2, group='fpuc', y='pov_gap_pc',
yaxis_title='Change in poverty gap',
title='Poverty gap by reform')
Inequality¶
Moving beyond poverty, we found that FPUC would have reduced overall inequality (as measured by the Gini index over individuals, considering their SPM unit’s total resources) more than the universal payments, by about 20 percent. The payroll tax cut would not have affected inequality.
line_graph(pov_gap_ineq2, group='fpuc', y='gini_pc',
yaxis_title='Change in Gini index',
title='Inequality by reform')
Moving forward¶
While FPUC slightly outperformed universal transfers in the baseline case, that trend strongly reverses when evaluating a FPUC expansion, conditional on FPUC being in effect from April to July. The difference is larger in worse economic times (2009 vs. 2018), when using the poverty gap measure to consider poverty depth, and is more pronounced for poverty of children, Black people, and especially people with disabilities.
The intuition behind these trends is that FPUC focuses resources on unemployed people, especially long-term unemployed. While these can be similarly poor to people not in the labor force, who constitute about half of the population in poverty, the first FPUC round helped this targeted group substantially. To reduce overall poverty, future resources are better spread out than adding more eggs to the unemployed basket.
Poverty rate¶
Starting with the poverty rate, we find that extending FPUC would have reduced poverty by 5.5 percent in 2009, compared to a baseline where FPUC had already been enacted from April through July. However, a budget-neutral universal payment would have reduce poverty by 6.8 percent–24 percent more than extending FPUC. An adult-only transfer would be somewhere in the middle, and a payroll tax cut would reduce poverty by about a third as much as the FPUC extension.
Compared to FPUC extension, universal payments have a stronger relative poverty impact on children (36 percent larger in 2009), Black people (70 percent), and people with disabilities (96 percent); see Appendix A for details.
# Round for the hover text.
line_graph(pov_rates2[(pov_rates2.age_group == 'All') &
(pov_rates2.race == 'All') &
(pov_rates2.disability_filter == False)],
group='fpuc2', y='pov_pc',
yaxis_title='Change in SPM poverty rate',
title='Poverty reduction by reform')
Poverty gap¶
In all years, universal payments reduce the poverty gap at least 50 percent more than FPUC extension would. For example, in 2009, FPUC extension would have reduced the poverty gap by 4.4 percent, compared to 6.3 percent from a budget-neutral universal payment—a 44 percent larger effect.
The larger deltas with this measure, compared to the poverty rate measure, reveal that universal programs are especially effective at reducing deep poverty, conditional on additional unemployment benefits.
# Round for the hover text.
line_graph(pov_gap_ineq2, group='fpuc2', y='pov_gap_pc',
yaxis_title='Change in poverty gap',
title='Poverty gap by reform')
Inequality¶
While the initial FPUC reduced inequality more than universal payments in every year, the opposite is true in most years; for example, universal payments would have reduced inequality 14 percent more in 2009. Payroll tax cuts continue to leave inequality unaffected.
line_graph(pov_gap_ineq2, group='fpuc2', y='gini_pc',
yaxis_title='Change in Gini index',
title='Inequality by reform')
Conclusion¶
US lawmakers will consider a range of factors in drafting the next relief bill, from infection containment to work incentives to economic growth to income stability.
To the extent that distributional issues are included in this list—and we believe that they should be—our research shows that the bill should include direct payments. Unlike most OECD countries, unemployment benefits in the US are regressive, and while FPUC’s flat structure is more progressive than relative reforms like increasing the wage replacement rate, it remains less progressive than universal payments. Payroll tax cuts barely reduce poverty and do not affect inequality.
Means-tested payments would be even more progressive than universal ones, but targeting in emergency situations introduces its own challenges. The IRS issued Recovery Rebates largely based on 2018 tax returns, which in some cases reflected income over 2 years prior to the pandemic. Not only may those tax returns fail to represent families’ current financial situation, they may fail to represent the families themselves, if people have since married, divorced, or had children. Depending on one’s income in the past two years, the timing of 2019 tax filing could determine whether one received or didn’t receive a Recovery Rebate. Addressing some of these issues may require designing payment systems outside the IRS, but concerns about providing payments to high-income people are easily addressed by recouping universal payments in the tax code, where systems for complicated household and financial situations already exist.
FPUC’s first round significantly reduced poverty and inequality, but continuing that momentum will require diversifying the set of recipients. By including low-income workers and people outside the labor force, universal payments provide that diversification. Our research shows that a consequence of this broadening is substantially greater poverty and inequality reduction compared to FPUC.g
Appendix¶
Appendix A: Supplemental poverty charts¶
The following charts compare the poverty effect of extending FPUC with budget-neutral alternatives for children, Black people, and people with disabilities.
Poverty among children¶
In 2009, FPUC extension would have been half as effective at reducing child poverty as universal payments. Unsurprisingly, the adult-only restriction dramatically reduces the effectiveness.
line_graph(pov_rates2[(pov_rates2.age_group == 'Children') &
(pov_rates2.race == 'All') &
(pov_rates2.disability_filter == False)],
group='fpuc2', y='pov_pc',
yaxis_title='Change in child SPM poverty rate',
title='Child poverty reduction by reform')
Poverty among Black people¶
The poverty reduction series for Black people are more volatile than for other demographics. FPUC extension have roughly two-thirds the antipoverty effect as universal payments in 2009, but in 2015 FPUC extension would have been 25 percent more effective, and by 2018, FPUC extension is again less effective.
line_graph(pov_rates2[(pov_rates2.age_group == 'All') &
(pov_rates2.race == 200) &
(pov_rates2.disability_filter == False)],
group='fpuc2', y='pov_pc',
yaxis_title='Change in Black SPM poverty rate',
title='Poverty reduction among Black people by reform')
Poverty among adults with disabilities¶
Adults with disabilities benefit more than other demographics from replacing an FPUC extension with universal payments: in 2009, extending FPUC would have reduced their poverty by 2.4 percent, compared to 5.5 percent from universal payments and 6.0 percent from adult-only payments—over double the effect.
Since the CPS only collects disability status for civilians aged 15+, we report poverty status for adults (aged 18+) with disabilities. This explains the strong performance of adult-only payments.
line_graph(pov_rates2[(pov_rates2.age_group == 'Adults') &
(pov_rates2.race == 'All') &
(pov_rates2.disability_filter == True)],
group='fpuc2', y='pov_pc',
yaxis_title='Change in PWD SPM poverty rate',
title='Poverty reduction among adults with disabilities by reform')
- 1
Assumes that workers have worked full-time at a constant wage for the past several quarters.
- 2
We also considered an imputation that assumes continuous unemployment spells starting in a random week between 1 and (52 - weeks unemployed). However, this approach produced larger costs for the 17-week FPUC than the 22-week FPUC extension, since the latter occurred at the end of the year and therefore provided fewer opportunities for workers to benefit from it, if their unemployment spell began earlier. In this alternate approach, universal payments reduced poverty more, relative to FPUC and FPUC extension.
- 3
Our code for tax unit creation was based on similar code by Ernie Tedeschi, Sam Portnow, and the Policy Simulation Library’s Tax-Data project, which creates tax units from non-IPUMS CPS records.
- 4
We need the tax liability for each person because some tax units include people from multiple SPM units.