Rethinking The Welfare State August 2023

Rethinking the Welfare State
Nezih Guner, Remzi Kaygusuz and Gustavo Ventura

August 2023
Abstract
The U.S. spends signi…cant amounts on non-medical transfers for its working-age
population in a wide range of programs that support low and middle-income house-
holds. How valuable are these programs for U.S. households? Are there simpler, welfare-
improving ways to transfer resources that are supported by a majority? What are the
macroeconomic e¤ects of such alternatives? We answer these questions in an equi-
librium, life-cycle model with single and married households who face idiosyncratic
productivity risk, in the presence of costly children and potential skill losses of females
associated with non-participation. Our …ndings show that a potential revenue-neutral
elimination of the welfare state generates large welfare losses in the aggregate, although
most households support the move as losses are concentrated among a small group. We
…nd that a Universal Basic Income program does not improve upon the current sys-
tem. If instead per-person transfers are implemented alongside a proportional tax,
a Negative Income Tax experiment, it becomes feasible to improve upon the current
system. Providing per-person transfers to all households is costly, and reducing tax
distortions helps to provide for resources to expand redistribution.
JEL Classi…cations: E62, H24, H31
Key Words: Taxes and Transfers, Universal Basic Income, Household Labor Supply,
Income Risk, Social Insurance
Guner, CEMFI, Spain (email: nezih.guner@cem….es); Kaygusuz, Department of Economics,

Durham University, UK and Faculty of Arts and Social Sciences, Sabanci University, Turkey (email:
[email protected]); Ventura, Department of Economics, Arizona State University, USA (email:
[email protected]). Guner acknowledges support from EU 7th Framework Collaborative Project
Integrated Macro-Financial Modeling for Robust Policy Design (MACFINROBODS), Grant no. 612796,
the Spanish Ministry of Economy and Competitiveness Grants ECO2014-54401-P. We thank seminar and
workshop participants at Bank of Spain, BGSE Summer Forum (Income Dynamics and Family), Boĝaziçi
University, Collegio Carlo Alberto, EUI, Federal Reserve Banks of Atlanta and Minneapolis, Georgetown U.,
Helsinki School of Economics, IMF, LAEF-Santa Barbara, Macroeconomics of Inequality Conference (St.
Louis Fed), NBER Summer Institute (Macro Public Finance), Queen Mary U., SED Annual Meetings, SITE
(Macroeconomics and Inequality), and Universities of Barcelona, Bristol, Cambridge, Connecticut, Emory,
Geneva, Helsinki, Lugano, Purdue, Rice, Texas A&M, and Southern California. Francesco Chiocchio, Euge-
nio Renedo, Juan Carlos Ruiz Garcia and Micole De Vera provided excellent research assistance. The usual
disclaimer applies.
1
1 Introduction
In this paper, we focus on the set of means-tested government transfers available to house-
holds of working age in the United States. These transfers are sizable and cover a wide range
of programs and tax credit provisions. We refer to them as the welfare state for short. We
ask: to what extent households value the current welfare state in the U.S.? Are there sim-
pler, welfare-improving ways to transfer resources that are supported by a majority? What
are the macroeconomic e¤ects of switching to such alternatives?
Several observations motivate our work. First, the welfare state is far from insigni…cant:
excluding healthcare transfers (Medicaid), spending in all di¤erent programs add up to
nearly 2.5% of GDP.1 The rules and details of various programs are routinely discussed as
key in a¤ecting labor supply, inequality and well being in di¤erent ways. Hence, reforms
or expansions of the current scheme are expected to have signi…cant aggregate, distributive
and welfare e¤ects. Second, most households are potentially two-earner households.2 This
matters as current transfers depend critically on marital status/gender di¤erences and the
presence of children. Furthermore, households with two potential earners can cope with
labor market shocks better than single-person households. As a result, social insurance
and redistribution policy recommendations for an economy with two (potential) earners are
likely to be di¤erent than those for a single-earner economy. Lastly, marital status and
gender di¤erences are usually not considered in the analysis of tax and transfer policies. In
particular, di¤erences by marital status and gender in wage and earnings inequality over the
life cycle are typically ignored. In this paper, we …ll a void by providing a macroeconomic
analysis that considers all these aspects. We do so by developing an equilibrium framework
with uninsurable shocks, labor supply decisions in two-earner households, costly children,
and a detailed representation of taxes and transfers.
We build an equilibrium life-cycle model with a number of novel features. First, we
introduce a rich degree of heterogeneity in our model economy. Individuals di¤er by skill
(i.e., education levels), gender, and marital status. Skilled and unskilled individuals face
distinct wage rates and di¤er on how fast their skills evolve as they age. In addition, single
and married individuals face permanent shocks at birth and uninsurable persistent shocks
1
To place this number in international perspective, note that OECD (2019) calculates that income support
to working age population as a fraction of GDP was 1.9% in the US. The numbers for several European
countries are much higher: Germany (3.5%), France (5.4%), Belgium (7.5%).
2
More than 60% of the U.S. labor force between ages 25 and 54 is married (Current Population Survey,
2000-2018).
2
over their life cycle. Second, we allow for labor-supply decisions of spouses at the extensive
and intensive margins. Third, in line with data, we jointly account for the presence of
children across married and single households, the timing of their arrival, and the associated
childcare costs. In particular, we account for the level and variation of childcare costs over
the life cycle as crucial determinants of female labor supply. Finally, we model the dynamic
costs and bene…ts of participation decisions by allowing the labor market skills of females to
depreciate due to childbearing disruptions.
Our parameterized model takes into account the di¤erent programs that comprise the
U.S. welfare state and the progressive income tax system, excluding healthcare transfers,
e.g., Medicaid and Medicare. Transfers in the model economy consist of three main compo-
nents. The …rst is the Earned Income Tax Credit that provides a refundable tax credit to
households with earnings. The second component relates to child-related transfers, e.g., the
Child Tax Credit and childcare subsidies. The last part consists of the means-tested trans-
fers, which are typically identi…ed with the "welfare" system in the U.S., e.g., Temporary
Assistance to Needy Families and Food Stamps. How much transfers households receive from
di¤erent programs crucially depends on their marital status, earnings, number of children,
and childcare expenses, and this dependence motivates our modeling choices. As such, a
detailed description of the welfare state is a crucial input in the analysis. Any reform creates
winners and losers, and the magnitude of these gains and losses critically depends on who
bene…ts from the current system.
Given the welfare state and the tax system, we parameterize our model using U.S. ag-
gregate and cross-sectional data. Our model economy is in line with how earnings inequality
evolves over the life cycle (by gender, skill, and marital status), the levels and life-cycle
changes in married females’participation rates, the life-cycle patterns of the gender wage-
gap, and the rise in consumption dispersion with age. Altogether, our model economy
presents a comprehensive macroeconomic model suitable to address the role and reforms of
the welfare state.
Findings We conduct three sets of experiments. First, we consider the hypothetical

complete elimination of the welfare state and concomitantly reduce the income taxes for all
households to achieve budget balance. This allows us to gauge the aggregate e¤ects of the
welfare state, and the valuation of the welfare state vis-a-vis a reduction in the tax burden.
Overall, eliminating the welfare state leads to an increase in hours worked and participation
3
rates of married females of about 3% and 4.5%, respectively, and an increase in output of
about 1.7%. We …nd that eliminating the welfare state leads to a sharp aggregate welfare loss
measured by a consumption compensating variation, of about 3.2% for a newborn individual
under the veil of ignorance. Quite interestingly, a substantial majority of newborns support
the hypothetical elimination of the welfare state (about 60.7%). This re‡ects the targeted
nature of the current system, which is highly valuable to poor households and in particular
to poor single mothers with children, while the majority of households either do not bene…t
from it or do so marginally.
We then introduce two major reforms to the welfare state. First, we replace the entire
welfare state with a single transfer per person. We dub this case a Universal Basic Income
(UBI). We search across steady states for the level of the transfer and the level of taxation
that maximize ex-ante welfare (under the veil of ignorance) that keeps the budget balanced.
We …nd that a generous transfer per person of about 3.2% of mean household income (about
$3,140 per person or $12,550 for a family of four in 2019 dollars) maximizes the welfare of
newborns.3 However, even this welfare-maximizing level of transfers leads to an aggregate
welfare loss of 1.3%. i.e., there is no UBI program that can improve upon the current system.
If we introduce a UBI scheme on top of the current welfare state, as most proponents of a
UBI advocate, only small transfers lead to welfare gains. A relatively small transfer of just
1% of household income leads to welfare losses with a majority of individuals against such a
program. Overall, our …ndings indicate that a UBI scheme is hardly a good idea in welfare
terms.
Second, we replace all transfers and current income taxes with a single transfer per
person and a proportional tax rate. We dub this case a Negative Income Tax (NIT). This
case then combines a drastic transfer reform with a drastic tax reform. Similarly to the UBI
case, we search across steady states for the level of the transfer and the associated tax rate
that maximize the ex-ante welfare of newborns and satisfy budget balance. We …nd that
a generous transfer of about 4.8% of mean household income (about $4,700 per person or
$18,800 for a family of four in 2019 dollars) maximizes ex-ante welfare with a gain of 0.2%
and leads to strong majority support among newborns (about 68.2%). If a reform allows
NIT transfers to di¤er between single and married households with more generous payments
to singles, the welfare gains are larger (0.7%) and the program still has majority support of
newborns (51.4%). The desirability of the NIT scheme becomes stronger when we take into
3
The mean household income in 2019 was about $98,000.
4
account transitions across steady states.
The upshot for the relative success of an NIT scheme is that a larger degree of redistrib-
ution is feasible given the smaller tax distortions that ensue in this case. As tax distortions
are reduced with a proportional tax, the size of the aggregate economy grows alongside the
needed tax revenue to …nance larger transfers. Therefore, an NIT scheme makes higher de-
grees of redistribution feasible. We show that the desirability of an NIT scheme is resilient
to variations in the environment.
In placing our …ndings in perspective, one important case that we consider is an NIT
reform in an economy with a di¤erent underlying inequality. The US economy has changed
signi…cantly during recent decades, with a sustained increase in inequality. Meanwhile, U.S.
households changed with rising educational attainment, higher female labor force participa-
tion, and greater assortative mating. How do these changes a¤ect the value of transfers for
all and the desirability of an NIT reform? To answer this question, we recalibrate our model
to salient features of the US economy in the 1980s and introduce a NIT reform. We …nd
that the NIT reform leads to even more signi…cant welfare gains in such an environment.
With less inequality, there is a lower demand for transfers and the welfare-maximizing NIT
transfer becomes smaller. As a result, the required proportional tax rate is also smaller and
larger welfare gains are possible. In the 1980 economy, 80% of newborn households favor the
NIT scheme, while the support in the baseline is 68.2%. Thus, introducing a NIT scheme
in an economy with today’s characteristics is more di¢ cult since the demand for transfers is
greater, and …nancing such larger transfers requires larger tax rates.
Related Literature Our paper is closely related to the literature that studies the
welfare and aggregate e¤ects of taxes and transfers in dynamic, general-equilibrium models
with heterogeneous agents. Recent papers in this literature include Guner, Lopez-Daneri
and Ventura (2016), Heathcote, Violante and Storesletten (2017), Badel, Huggett and Luo
(2020), Kindermann and Krueger (2020), Boar and Midrigan (2022), and Ferriere, Grubener,
Navarro and Vardishvili (2023). Within this literature, Ferriere et al. (2023) emphasize how
larger transfers can be …nanced by lower progressivity of income taxes. Kaygusuz (2010,
2015), Guner, Kaygusuz and Ventura (2012, 2020), Ortigueira and Siassi (2013), Holter,
Krueger, and Stepanchuk (2019), Wu and Krueger (2021), and Borella, De Nardi and Yang
(2023), among others, consider environments with two-earner households. Ortigueira and
Siassi (2022) study how transfers can a¤ect cohabitation vs. marriage incentives and Low,
5
Meghir, Pistaferri and Voena (2022) analyze how marriage prospects can impact decisions
to participate in programs with time limits. Blundell, Pistaferri, and Saporta-Eksten (2016)
provide empirical evidence on the importance of family labor supply for consumption smooth-
ing.
The UBI and its close-cousin NIT have a long intellectual history (Mo¢ tt, 2003a) and
gained support in recent public debate. Van Parijs and Vanderborght (2017) and Hoynes and
Rothstein (2019) provide excellent reviews. Within macro-public-…nance literature, Lopez-
Daneri (2016) …nds that an NIT transfer of about 11% of mean income leads to a large,
2.1%, welfare gain despite sharp output losses. Luduvice (2019) and Conesa, Li and Li
(2023) consider replacing current transfers with a UBI and …nd that welfare gains are hard
to achieve, as we …nd in this paper. Using search and matching models, Jaimovich et al.
(2021) also …nd that UBI implies welfare losses, while Rauh and Santos (2022) suggest that
welfare gains are possible if UBI also replaces the current unemployment bene…ts. Daruich
and Fernandez (2023) study a UBI experiment within an overlapping generations model
where the next generation’s human capital depends on parents’decisions and …nd that UBI
is not a good idea when the welfare of future generations is taken into account.
Our analysis di¤ers from these papers on three key aspects. First, we provide novel facts
on how inequality along the life cycle changes for individuals and households of di¤erent
marital status and skill levels and use them to discipline the benchmark economy. Second,
the model economy features a comprehensive welfare state, necessary to identify winners
and losers in any reform. Finally, the model economy consists of single and married house-
holds, and married females who make participation decisions. These features are critical to
understanding the implications of any reform to the current transfer system since female la-
bor supply responds signi…cantly to changes in tax-transfer policies, and the current welfare
system treats di¤erent households (married/single, with and without children) di¤erently.
The paper is organized as follows. In section 2, we document patterns of hours, earnings
and consumption over the life cycle of individuals and household in the United States. Section
3 presents the model economy. In section 4 we describe the parameterization and calibration
of the benchmark economy. Section 5 discusses the properties of the benchmark economy.
In section 6, we present the main …ndings of our quantitative experiments. Section 7 places
our …ndings in perspective. Section 8 concludes.
6
2 Earnings, Hours and Consumption: Life-Cycle Facts
We use the March Supplement of the CPS from 1980 to 2019 to document how average
hourly wages, inequality of hourly wages and earnings, and labor market statistics (hours and
participation) change over the life cycle. For age pro…les for nondurable consumption, we use
the Consumption Expenditure Survey (CEX) from 1984 to 2019. Our measure of inequality
is the variance of logs. The Appendix provide sample restrictions and the de…nitions of all
the variables.
We estimate age pro…les using repeated cross sections in the data. To this end, let mj;t;c
be any statistic of interest for an age-j individual (or household) at time t; of cohort c.
For example, mj;t;c could be the variance of log hourly wages among j = 30 year olds in
2000, who are born in c = t j = 1970; i.e. the variance within a (j; t; c)-cell. Since age,
time and cohort are linearly dependent, we construct age pro…les using two approaches. We
…rst consider a time-e¤ects speci…cation by regressing mj;t;c on a set of age and time (year)
dummy variables, i.e.,
0 0
mj;t;c = j Dj + t Dt + "j;t;c ; (1)
where Dj and Dt are a set of age and time dummies. The underlying assumption in the
time-e¤ects speci…cation is that changes in mj;t;c over time are due to time-varying factors
that a¤ect every age (cohort), and once we control for time e¤ects we recover the age pro…les.
Equation (1) is estimated separately for each gender (men and women), marital status (mar-
ried and single), and skill group. For skills, we divide individuals in two groups; skilled (s),
those with at least four years of college education or more, and unskilled (u), with strictly
less than college education. The age pro…les are given by the estimated j values. Then, we
also estimate a cohort-e¤ects speci…cation, given by
0 0
mj;t;c = j Dj + c Dc + j;t;c ; (2)
where Dc is a set cohort dummies. In contrast to equation (1), the underlying assumption in
the cohort-e¤ects speci…cation is that changes in mj;t;c over time re‡ect di¤erences between
younger and older cohorts.
The assumptions behind these two speci…cations might be too strong for any speci…c age
pro…le we construct in this section. In our benchmark analysis, we use the time-e¤ects spec-
i…cation, which accounts better for observed trends in inequality (Heathcote, Storesletten,
7
and Violante, 2005). As a result, for all age pro…les we document in this section, we assume
that changes over time are captured by time e¤ects that in‡uence all ages at a given time.
For some variables, e.g. female labor force participation (Goldin, 2021), this assumption
might be restrictive, and a cohort perspective might be more appropriate. In section 7 and
in the Appendix, we present our analysis and the pro…les with cohort e¤ects.
The key …ndings that emerge from the benchmark analysis are listed below.
1. For males, as it is well known in the literature, the variance of log-hourly wages increase
non-trivially along the life cycle; see Figure 1 (left panel). As dummies, estimated j values
only capture variance of log wages for each age relative to an omitted one. Therefore, we
normalize 25 to its value in the data (the variance of log wages for 25 years olds, averaged
across years) and rescale all other coe¢ cients accordingly. The increase is more pronounced
for skilled than for unskilled men. The increase is of nearly 30 log points for married skilled
men between ages 25-60, versus a corresponding increase of about 14 log points for married
unskilled men. These patterns hold for single men as well, albeit with a smaller increase in
variances over the life cycle. These patterns are mirrored when inequality in labor earnings
rather on hourly wages is considered.
Single Males Married Males Single Females Married Females

0.40
0.40
0.50
0.50
0.35
0.35
0.40
0.40
Variance of Log Wages

0.30
0.30
0.30
0.30
0.25
0.25
0.20
0.20
0.20
0.20
25 30 35 40 45 50 55 25 30 35 40 45 50 55 25 30 35 40 45 50 55 25 30 35 40 45 50 55
Age Age Age Age
Unskilled Skilled Unskilled Skilled Unskilled Skilled Unskilled Skilled
Figure 1 - Variance of Log Wages, Males (left) and Females (right)
2. For females, married or single, we do not observe a similar increase. This is largely
independent of marital status and skill –see Figure 1 (right panel). The increase in dispersion
in hourly wages for unskilled (skilled) females is of about …ve (ten) log points up to age 40,
and after that, the level of dispersion is roughly constant. This is in stark contrast with the
8
increase in dispersion for males discussed in point 1 above.4
3. For both married and single households, the variance of log earnings increase non-
trivially along the life cycle, but the level of inequality is much lower among married house-
holds. At age 25 (45), variance of log earnings is about 0.38 (0.55) for all households, but
only 0.30 (0.41) for married households.
4. The gender wage gap, de…ned as the ratio of average hourly earnings of females relative
to males, increases over the life cycle. These changes are sharper for skilled individuals, with
a decline in this ratio from about 92% at age 25 to about 70% at age 45. For unskilled
individuals, the corresponding change is smaller and of about 11 percentage points. Figure
2 (left panel) displays these patterns.
5. Over the life cycle, the participation rate of married females …rst declines and then
rises up to ages 45-48, and then declines again. These changes are much more pronounced
for married skilled females. Figure 2 (right panel) displays these patterns.
Gender Gap Married Female LFP

1.00
0.85
0.80
0.90
Labor Force Participation

Gender Gap
0.75
0.80
0.70
0.70
0.65
0.60
25 30 35 40 45 50 55 25 30 35 40 45 50 55
Age Age
Unskilled Skilled Unskilled Skilled
Figure 2 - The Gender Wage Gap (left ); LFP of Married Females (right)
6. Conditional on work, there is signi…cant variation in hours of work among married

females, measured by the variance of log hours at each age. The level is, nevertheless, roughly
constant over the life cycle, at around 0.13; see Figure 3 (left panel).
7. The correlation between earnings of husbands and wives is low, around 0.15 at ages
40-50, and slightly \-shaped early in the life–cycle. Figure 3 (right panel) displays these
patterns.
4
Bayer and Kuhn (2020) document similar gender di¤erences in life-cycle pro…les of earnings inequality
in Germany.
9
Var of Log Hours Correlation of Earnings
0.20
0.30
0.25
0.15
0.20
Variance of Log Hours
Correlation
0.10
0.15
0.10
0.05
0.05
0.00
0.00
25 30 35 40 45 50 55 25 30 35 40 45 50 55
Age Age
Figure 3 - Var. of Log Hours, Married Females (left); Correlation of Spousal Earnings (right)
8. The variance of log consumption increases along the life-cycle, but much less than
the increase in the variance of household or individual earnings. The increase peaks at age
55, about 0.12 log points above its level at age 25. This is a well-known fact by now, and
documented in Aguiar and Hurst (2013) and Primiceri and van Rens (2009), among others.
3 The Economic Environment

We study a stationary life-cycle economy populated by a continuum of males (m) and a
continuum of females (f ). Let j 2 f1; 2; :::; Jg denote the age of each individual. Each
model period is one year, and the …rst model period corresponds to age 25. Population
grows at rate n: The life-cycle of agents is split into two parts. Each agent starts life as a
worker and at age JR ; individuals retire and collect pension bene…ts until they die at age J:
Individuals di¤er in their marital status. We assume that they are born as either single
or married and their marital status does not change over time. Each individual is also born
with a given intrinsic type (education) that de…nes the rental rate for his/her labor services,
and the growth of their labor endowment as they age. Married households are comprised by
individuals who are of the same age.
Married households and single females also di¤er in terms of the number of children
attached to them. They can be childless or endowed with children. The number of children
depend on the educational attainment of the parents and they appear either early or late in
the life-cycle exogenously. Children a¤ect the resources available to households for several
periods, and this is mitigated partially or fully by government policies targeted to children.
Children do not provide any utility.
10
Individuals also di¤er in terms of permanent shocks received at the start of life, which
are correlated among spouses. Furthermore, each period, individuals experience uninsurable
productivity shocks, which a¤ect how much they can earn per hour. We assume that these
shocks are persistent. We also assume that shocks that husbands and wives receive are
correlated. Hence, heterogeneity among households arises due to di¤erent factors; their
education level, the permanent and life-cycle shocks of their members, and who is married
with whom. These forms of ex-post and ex-ante heterogeneity determine, in conjunction with
labor supply and savings decisions, the degree of income, consumption and wealth inequality
in the economy.
Production and Markets There is an aggregate …rm that operates a constant re-
turns to scale technology. The …rm rents capital, skilled and unskilled labor services from
households at the rates R, ws and wu , respectively. Using K units of capital and L units of
the composite labor input, the …rm produces
1
F (K; L) = K L1 ; with L Ls + (1 )Lu;g ; 2 ( 1; 1);
where Ls and Lu;g stand for the stock of skilled labor, and unskilled labor used in the
production of goods, respectively. The elasticity of substitution between labor of di¤erent
1
types is constant and given by = 1
:
We assume that capital depreciates at rate k. Childcare services are provided using
unskilled labor services only. Thus, the price of childcare services is the wage rate, wu . As
a result, unskilled labor services available are split between the production of consumption
and investment goods, Lu;g , childcare services, Lu Lu;g : Households save in the form of a
risk-free asset that pays the competitive rate of return r = R k.
Ex-ante Heterogeneity and Demographics At the start of life, each male is en-
dowed with an exogenous type z that remains constant over his life cycle: z 2 Z = fu; sg.
This type of heterogeneity de…nes whether the agent is skilled (s) or unskilled (u) that we
later map to educational levels in the data. For females, we equivalently have x 2 X = fu; sg.
We assume that each cohort is 1 + n bigger than the previous one. These demographic pat-
terns are stationary so that age-j agents are a fraction j of the population at any point in
time. The weights are normalized to add up to one, and obey the recursion, j+1 = j =(1+n):
11
3.1 Labor E¢ ciency Units
We consider a general structure, where individuals di¤er at the start of the life cycle in
their skills, permanent shocks, as well as uninsurable shocks experienced as they age. These
shocks are dependent on the skill of individuals (u; s), their gender (m; f ) and their marital
status (M; S).
Singles Consider …rst single males. Their labor endowment (e¢ ciency units) at age j
is given by
S S
$m (z; j) exp( m;z + m;z;j ); z 2 Z = fu; sg;
where the function $m (:; :) summarizes the combined e¤ects of skill and age on the labor
endowment. is a permanent shock and is a persistent shock. We assume that the
S 2
permanent shock is normally distributed: m;z N (0; S
m;z
); z 2 Z:
We assume that for j > 1, the persistent shock is governed by a random walk, given by
S S
m;z;j+1 = m;z;j + "Sm;z;j+1 ; z 2 Z;
with "Sm;z;j+1 N (0; 2

"S
m;z
) representing innovations over time. We furthermore assume that
S
the initial value of at the start of the life cycle is zero; i.e., m;z;1 = 0; z 2 Z.
The structure is di¤erent for single females, as their e¢ ciency units evolve endogenously,
with growth and depreciation rates that depend on intrinsic skills and labor market experi-
ence. Intrinsic skills determine their initial human capital: h1 = $f (x; 1), x 2 X. For j > 1,
we have
h0 = H(x; h; l; e) = exp [ln h + e

x (l) x (1 (l))] ; x 2 X = fu; sg; (3)
where e stands for labor market experience and (:) is an indicator function that is 1 if hours
e
worked are positive and zero otherwise. The parameter x is the experience-skill growth rate
and x stands for the depreciation rate. It follows that for a single female of age-j who has
S S
human capital h; her realized labor e¢ ciency is given by h exp( f;x + f;x;j ): The permanent
and the persistent shock obey the same representation as for males, with innovation variances
that depend on marital status and skill.
12
Married Couples Married individuals draw permanent shocks at the start of their life
cycle that are potentially correlated. They also draw values for their persistent shocks which
are potentially correlated as well. The labor endowments (labor e¢ ciency) of a married male
and a married female are given by
M M M M
$(z; j) exp( m;z + m;z;j ) and h exp( f;x + f;x;j ); z 2 Z; x 2 X:
M M
The labor e¢ ciency of a married female is correspondingly given by h exp( f;x + f;x;j ),
where h follows the same law of motion for singles; equation (3).
M M M M
The initial conditions are such that m;z;1 = 0 and f;x;1 = 0: For j > 1, m;z;j and f;x;j
follow a bivariate process, given by
M M
m;z;j+1 = m;z;j + "M
m;z;j+1 and
M
f;x;j+1 = M
f;x;j + "M
f;x;j+1 for z 2 Z; x 2 X
with !
2
0 "M "f "m
("M M
m;z;j+1 ; "f;x;j+1 ) N ; m;z
2 ; z; x 2 Z X:
0 "f "m "M
f;x
The values of permanent shocks for married individuals are draws from a bivariate normal
distribution as well. That is,
2
!
M M 0 M
m;z f m
( m;z ; f;x ) N ; 2 ; z; x 2 Z X:
0 f m M
f;x
Note that we assume that while innovations depend on skills, the covariance structure for
both permanent and persistent shocks does not. This parsimonious speci…cation allows us
to capture key correlations across married spouses, both at the start as well as in along the
middle of the life cycle –see section 5.
Labor Earnings We now summarize the notion of labor earnings resulting from our
choices, taking into account skill prices (ws and wu ), endowments and labor supply choices
–described later. For an age-j single male of type z, earnings are given by
wz $(z; j) exp( Sm;z + S

m;z;j ) lm
|{z} | {z } |{z}
wage by skill labor e¢ ciency labor supply
For a single female of skill x 2 X who has human capital h; age j, earnings are given by
wx h exp( Sf;x + S
f;x;j ) lf
|{z} | {z } |{z}
wage by skill labor e¢ ciency labor supply
13
Finally, for a married couple of skill z; x 2 Z X; of age j, when she has h units of human
capital, earnings are given by
wx h exp( M
f;x +
M
f;x;j ) lf + wz $(z; j) exp( M + M
m;z;j ) lm
|{z} | {z } |{z} |{z} | {zm;z } |{z}
wage by skill labor e¢ ciency labor supply wage by skill labor e¢ ciency labor supply
3.2 Children and Childcare Costs

Children are assigned exogenously to married couples and single females at the start of life,
depending on the education of parents. Each married couple and single female can be of
three types: without any children, early child bearers, late child bearers. We denote this
dimension of heterogeneity by b = f0; 1; 2g:
If b 6= 0, children appear deterministically at parents’age j(x; z; b) for married households
and j(x; b) for single females. Married households have k(x; z) children, while single females
have k(x) children. For married households, half of the children appear at age j(x; z; b) and
the other half at age j(x; z; b) + 2; i.e., children are spaced by two years. It is equivalent
for single households: half of the children appear at age j(x; b) and the other half at age
j(x; b) + 2: Each child stays with their parents for N model periods.
We assume that if a female with children works, married or single, then the household
has to pay for childcare costs. Childcare costs depend on the age of the child, t; and are
priced at rate wu . We assume that children in single female households require d(x; t) units
of childcare services per child, t = 1; 2; :::; N: Married households require d(x; z; t) units of
childcare services per child. Since competitive price of childcare services is the unskilled
wage rate wu , the cost of childcare services per child equals wu d(x; t) for single females and
wu d(x; z; t) for married households.
3.3 Preferences
The momentary utility function for singles is given by
1
U S (c; l) = log(c) Bi (l)1+ ; i = m; f
where c is consumption, l is time devoted to market work, and is the intertemporal elas-
ticity of labor supply (Frisch elasticity). The parameter Bi captures potential gender-driven
di¤erences in the disutility of work.
Married households maximize the sum of their members utilities. We assume that when
the female member of a married household works, the household incurs a utility cost q:
14
We assume that at the start of their lives married households draw a q 2 Q; where Q
R++ is a …nite set. These values of q represent the utility costs of joint market work for
married couples. For a given household, the initial draw of utility cost depends on the type
(education) of the husband. Let (qjz) denote the probability that the cost of joint work is
P
q, with q2Q (qjz) = 1. We assume that for married households with children at home,
the utility cost q is multiplied by a factor that depends on the age of the youngest child at
home, tmin and the mother’s skill level, #x (tmin ), x 2 X: This speci…cation captures the idea
that joint work becomes more costly with arrival of children, beyond childcare costs, and
that this additional cost changes as children grow older.
Formally, if b 2 f1; 2g and the household age is such that j(x; z; b) j j(x; z; b)+N +2,
i.e., children are at home (recall that the …rst child arrives at j(:) and the second one leaves
at j(:) + N + 2), then the period utility of a married household is given by
1+ 1 1+ 1
U M (c; lf ; lm ; ; q; j) = 2 log(c) Bm lm Bf lf flf gq(1 + #x (tmin )): (4)
where f:g denotes the indicator function.5 For households without any children at home,
#x (tmin ) = 0:
Note that consumption is a public good within the household. The variable captures
heterogeneity in the disutility of work across married females. We assume that is realized
at the start of life, and takes two values with equal probability; 2 f L ; H g: Note also
that the parameter > 0, the intertemporal elasticity of labor supply, is common for all
individuals; males or females, married or single. It is also important to note that following
the tradition in macroeconomics literature, we restrict the preferences to be consistent with a
balanced-growth path. As in, e.g., Attanasio et al. (1999, 2005), we could allow the marginal
utility of consumption to be a¤ected by the female labour force participation decision. In the
current speci…cation, the female labour force participation and demographics (the number
of children) a¤ect the level of utility through the cost of joint work.
3.4 Taxes and Transfers

There is a government that taxes labor and capital income, and uses tax collections to pay for
government consumption, tax credits, transfers to individuals. It also runs a pay-as-you-go
social security system, so it collects payroll taxes and pays retirement bene…ts.
5
Note that if x, z and j are known, the age of the youngest child can be readily calculated.
15
Transfers Households in the model have access to transfers that depend on gender,
marital status and household income. Income for tax and transfer purposes is labor plus asset
income. For a household with income level I, number of children k; childcare expenses D; the
S
transfers are represented by functions T RfS (I; k; D); T Rm (I) and T RM (I; k; D), for a single-
female, single-male and married-couple households, respectively. This generic formulation
of transfers allows us to capture a host of transfers and tax credit programs in the United
States. We describe below how these functions are parameterized in light of data.
Taxation and Social Security The total income tax liabilities of married and single
households, before any tax credits, are a¤ected by the presence of children in the household,
and are represented by tax functions T M (I; k) and T S (I; k), respectively, where k stands for
the number of children at the household. These functions are continuous in I, increasing
and convex. This representation captures the e¤ective variation in tax liabilities associated
to income, marital status and the presence of children in households.
There is a (‡at) payroll tax that taxes individual labor incomes, represented by p, to fund
social-security transfers. Moreover, each household pays an additional ‡at capital income tax
for the returns from his/her asset holdings, denoted by k. Retired households have access
to social security bene…ts. The social security bene…ts depend on agents’education types,
i.e., initially more productive agents receive larger social security bene…ts. This allows us
to capture in a parsimonious way the positive relation between lifetime earnings and social
security transfers, as well as the intra-cohort redistribution built into the system. Let pSf (x);
pSm (z); and pM (x; z) indicate the level of social security bene…ts for a single female of type
x, a single male of type z and a married retired household of type (x; z), respectively. The
social security system has to balance its budget every period.
3.5 Decision Problem

We now present the decision problem for di¤erent types of agents in the recursive language.
We provide a formal de…nition of a stationary equilibrium in the Appendix. We focus on
single females and married couples, since the problem of single males is rather standard. For
ease of notation, the dependence of shocks on type, gender and marital status is suppressed
S S
whenever possible. For single females, the individual state is (a; h; e; x; f;x ; f;x ; b; j); where a
stands for asset holdings. For married couples, the state is given by (a; h; e; x; z; ; M M M
m;z ; f;x ; m;z ;
M
f;x ; q; b; j).
Note that the dependency of transfers and taxes on the presence of children in the house-
16
hold is summarized by age of parents (j) and childbearing status (b), in conjunction with
x for single females and the pair (x; z) for married couples. The same reasoning applies for
childcare costs, or the utility costs of joint participation for married couples when children
are present. That is, if we know the intrinsic type of a single female or a married household,
the age of parents (j) and fertility type (b), we know the age of each child and the childcare
costs. Given parents’types, the half of children appear at parents’age j(:) and the other
half at j(:) + 2: Then, when their parents are of age j; young and old children at home have
ages j j(:) + 1 and j j(:) + 3.
For expositional purposes, we collapse the permanent/exogenous characteristics in the
household problems in a single vector of state variables. For single females, let SfS
S
(x; f;x ; b) be the vector of variables that do not change along the life-cycle for single females
and single males, respectively. For married households, let S M (x; z; ; ; q; b) be the vector
M M
of such states for married households, with ( f;x ; m;z ). In similar fashion, for the case
M M
of married couples, we summarize the pair of persistent shocks by ( f;x ; m;z ). Likewise,
for expositional purposes, we denote by EfS (x; h; S
f;x ; ; lf ) and E M (x; z; h; ; ; lm ; lf ; j), the
labor earnings of single females and married couples, respectively, as de…ned in Section 3.1.
The Problem of a Single Female Household Given her current state, (a; h; e; SfS ; ; j)
the problem of a single female is
VfS (a; h; e; SfS ; ; j) = max

0
fU S (c; l) + E 0j VfS (a0 ; h0 ; e0 ; SfS ; 0 ; j + 1)g;
a ;l
subject to
(i) With kids: if b = f1; 2g; j 2 fj(x; b); j(x; b) + 1; :::; j(x; b) + N + 2g
S
a(1 + r(1 k )) + Ef (x; h; ; ; l)(1 p)
c + a0 = S S ;
+T Rf (I; K; D) T (I; K) wu D (l)
where I = EfS (x; h; ; ; l) + ra. K is the number of children present in the household, either
old, born at j(x; b), or young, born at j(x; b) + 2. It is given by
2 3
k(x; b) 4
K= (j(x; b) j j(x; b) + N ) + (j(x; b) + 2 j j(x; b) + 2 + N )5 :
2 | {z } | {z }
old children young children
Meanwhile, D stands for the childcare expenses incurred:

k(x; b)
D = d(x; j j(x; b) + 1) (j(x; b) j N) +
2
k(x; b)
d(x; j j(x; b) + 3) (j(x; b) + 2 j j(x; b) + 2 + N ):
2
17
(ii) Without kids but not retired: if b = 0, or b = f1; 2g and j 2
= fj(x; b); :::; N + 2g, then
there are no children at home and
c + a0 = a(1 + r(1 k )) + EfS (x; h; ; ; l)(1 p)
+T RfS (I; 0; 0) T S (I; 0):
iii) Retired: if j JR , then there are no children and
c + a0 = a(1 + r(1 k )) + pSf (x) T S (ra; 0) + T RfS (ra; 0; 0):
In addition,
x
h0 = H(x; h; lf ; e) = exp [ln h + x
e (lf ) (1 (lf ))] ; (5)
e0 = e + (l) and l 0, a0 0 (with strict equality if j = J + 1):
The Problem of Married Households Like singles, married couples decide how
much to consume, how much to save, and how much to work. They also decide whether the
female member of the household should work, taking into account the evolution of her skills,
experience and childcare costs. Note that in the formulation below, we make the current
utility of married households to depend on (x; z; b; j); as these variables fully determine the
age of children present in the household that may a¤ect the disutility of joint market work,
q(1 + #x (tmin )) term above. Formally, the problem is given by
V M (a; h; e; S M ; ; j) = 0max fU M (c; lf ; lm ; q; x; z; b; j) + E 0j V M (a0 ; h0 ; e0 ; S M ; 0

; j + 1)g;
a ; lf ; lm
subject to
(i) With kids: if b = f1; 2g; j 2 fj(x; b); :::; N + 2g, then
M
a(1 + r(1 k )) + E (x; z; h; ; ; lm ; lf ; j)(1 p)
c + a0 = ;
T M (I; K) + T RM (I; K; D) wu D (l);
where I = E M (x; z; h; ; ; lm ; lf ; j) + ra; and lm 0, lf 0; and a0 0:

In this formulation, E 0j now represents the joint expectation over the shocks that hus-
bands and wives face. The number of children present, K; and childcare expenses are D are
formulated as they were done for a single female. The household problem also takes into
account the accumulation of human capital for the wife, given by h0 = H(x; h; lf ; e) and
e0 = e + (l): The budget constraints when the household is not retired but without any
children and when the household is retired, cases (ii) and (iii), are de…ned accordingly.
18
3.6 Sources of Inequality in the Model
What are the determinants of inequality at a point in time and over the life cycle across
individuals and households in the model? This question is of central importance in assessing
the e¤ects of transfer policies.
First, individuals di¤er in their intrinsic skills and that experience permanent and per-
sistent shocks at birth. Permanent and persistent shocks are common in life-cycle models
with heterogeneous individuals. Di¤erent from most of the work in the area, di¤erences in
skill type at birth determine (i) potentially di¤erent growth rates in labor productivity be-
tween skilled and unskilled individuals, and (ii) between-group di¤erences as individuals face
di¤erent rental rates for labor services depending on their skill type. Point (i) implies that
our model encompasses a mixture of traditional parameterization of heterogeneity (usually
referred to as Representative Income Processes or RIP), with a human capital view of dif-
ferences of individuals as they age, as emphasized in Guvenen (2009) and Huggett, Ventura
and Yaron (2011), among others (Heterogeneous Income Processes or HIP).
The second layer of heterogeneity determining inequality concerns marital status. At
birth, some individuals are single, some are married, and married ones are assigned to spouses
according to their skill type. Besides, within a given skill pair, permanent and persistent
shocks are potentially correlated between spouses. Overall, as in Greenwood et al. (2016)
and others, marriage can amplify existing di¤erences between individuals and contribute to
propagating shocks over the life cycle.
Finally, di¤erences in individuals by gender, coupled with children’s presence, help de-
…ne the level of gender premia in wages at birth and its evolution over the life cycle. As
children appear and women leave the workforce, skill depreciation kicks in, and thus, the
gender gap in wage rates grows over time. As children require fewer resources as they age,
some women return to work, accumulate skills again and the gender-wage gap moderates its
growth. As we describe below in our analysis of the benchmark economy, women’s behavior
regarding participation over time, in conjunction with uninsurable shocks, determines gender
di¤erences in the life-cycle pro…le of earnings inequality.
The reforms of the welfare state that we consider, the Universal Basic Income (UBI)
and Negative Income Tax (NIT), have simple structures. However, they replace the existing
welfare state, which is not simple at all. As a result, identifying the winners and losers re-
quires rich ex-ante heterogeneity, since di¤erent socio-economic groups receive quite di¤erent
transfers in the benchmark economy.
19
3.7 Modelling Choices
Several model elements are taken as exogenous in the current analysis. This allows us to
simulate a model with extensive heterogeneity in educational attainment, marital status
and sorting, and the number and timing of children. As we emphasize in the paper, such
heterogeneity is crucial to understanding the welfare consequences of reforms to the welfare
state. We model these features as exogenous, which allows us to focus on a subset of critical
endogenous decisions: household labor supply, female human capital accumulation, and
savings. There is an extensive empirical literature on the incentive e¤ects of the welfare state
on marriage and fertility, with a particular focus on the impact of the 1996 welfare reform.
The …ndings from this literature suggest that the incentive e¤ects of the welfare state on
marriage, fertility, and single motherhood are modest; see Bitler et al. (2004), Kearney
(2004), and Mo¢ tt et al. (2020). Hence, we expect that the welfare state’s direct impact on
marriage and fertility incentives likely to be small. On the other hand, it is well established
that children that grow with single mothers receive relatively much less investment than those
with two parents. As a result, even small adverse e¤ects on marriage incentives coupled with
adverse e¤ects on children can accumulate across generations, impacting intergenerational
mobility. This is emphasized by Aiyagari, Greenwood, and Guner (2000), among others.
We also have a unitary model of household decisions. With a non-unitary model like in
Voena (2015), transfers can a¤ect within-household allocations even if they do not change
marriage and divorce decisions. In particular, a non-unitary model could allow us to study
how gender-speci…c taxes and transfers can alter household allocations and within-household
distribution of welfare.
The model also abstracts from health shocks and government-provided health insurance
programs prior to retirement, such as Medicaid, Social Security Disability Insurance (SSDI)
and Children’s Health Insurance Program (CHIP). Overall, since 2015, expenditures on
these programs have averaged about 3.9% of GDP, with about 3.1% on Medicaid, 0.7%
on SSDI and about 0.1% on CHIP.6 Recent papers introduce exogenous health shocks into
heterogeneous-agent macro models – see Hosseini, Kopecky, and Zhao (2021) for a recent
example. In these models, health shocks a¤ect earnings (mainly through labor supply) and
act like income shocks, and as a result, government-provided health insurance is valued by
6
For expenditures on Medicaid and CHIP, see Centers for Medicare and Medicaid Services
(https://www.cms.gov/). For expenditures on disability, see Annual Statistical Supplements to the Social
Security Bulletin (https://www.ssa.gov/policy/docs/statcomps/supplement/).
20
individuals. Hosseini et al. (2021), for example, …nd that health shocks account for about
30% of lifetime earnings inequality, and government provided disability insurance is valued
by consumers. Since health shocks are exogenous in these models, we see our paper and
this literature as complementary, each focusing in detail on di¤erent aspects of the welfare
system.
4 Parameter Values
This section describes how we select parameter values to compute a stationary equilibrium.
We relegate details to the Appendix (Tables A11 and A12 summarize our parameter choices).
The model period is one year. Agents start their life at age 25, potentially work for forty
years, retire at age 65 (j = JR ); and then live until age 80 (j = J): The population grows
at the annual rate of 1.1%. Skilled individuals are those with at least four-year college
degree. The marital structure (who is single, who is married and who is married with
whom), childbearing status, and the number of children for di¤erent types of households are
taken directly from the data.
Endowments For males, following the procedure described in Section 2, we construct

age pro…les of mean hourly wages for each skill group using data from 1980-2019 CPS March
Supplement, and set $m (z; j), z = u; s, to these pro…les (Figure A7 in the Appendix).
For females, we use age-25 wage levels to calibrate their initial human capital levels, h1 =
$f (x; 1). After age 25, female skills evolve according to equation (3).
e
We select the parameter x so that if a type-x female works for one more period, her wage
grows exactly at the same rate as a male of the same type with the same experience level (e).
Hence, if a female works in every period, her labor market productivity evolves exactly like a
male, except for the observed age-25 wage gender gap. Figure A7 in the Appendix shows the
calibrated values for the growth factors. For depreciation rates, we select each one so that
the model is consistent with the evolution of the wage-gender gap for the …rst decade of the
life cycle (ages 25-35). The resulting values are u = 0:025; and a non-trivially higher value
for skilled females, s = 0:059. These values are required to reproduce the faster increase in
the wage-gender gap with age for skilled females documented in section 2.7
7
Blundell et al (2016) …nd similar results for the UK.
21
Productivity Shocks There are in total eighteen parameters that determine the pro-
ductivity shocks: eight variances for permanent shocks (by skill, gender, and marital status),
eight innovation variances for persistent shocks (again by skill, gender, and marital status,
plus two covariances (for permanent shocks and innovations to persistent shocks). Table A11
in Appendix presents these parameters. For permanent shocks ( ), we match the observed
variances of log-wages at age 25 by skill, gender and marital status. To pin down the value
of covariance term for married individuals, f m ; we target the correlation in log-wages
among all spouses at age 25. For the variances of innovations to persistent shocks ("), we
target the observed variances of log-wages towards the end of the life cycle (age 54) for each
group. For the covariance of innovation in persistent shocks across spouses, "f "m ; we target
the correlation of wages between husbands and wives by middle age (ages 40-45). Over-
all, the variances of innovations for persistent shocks for men are substantially larger than
for females, while the corresponding variances for skilled individuals, male or female, are
larger than for unskilled ones. Overall, not surprisingly, the innovation variances are smaller
than in related estimates, e.g., Heathcote, Storesletten, and Violante (2010) and Huggett
et al (2011). This re‡ects the division of individuals between skilled –who experience faster
growth in labor e¢ ciency with experience–and unskilled ones, as well as the distinction of
individuals by gender and marital status.
Income Taxation To compute the tax functions, i.e., T S (I; k) and T M (I; k); we adopt
a parametric form for the average tax rate:
(I) = 1 I ;
where I (income) is measured in multiples of mean household income and (I) is the average
tax rate. The parameter determines the progressivity of the tax scheme and determines
its level. The parameters and depend on marital status and the number of children,
and are estimated from IRS micro data on tax returns. Since the EITC, CTC and CDCTC
are explicitly modeled in the benchmark economy, the tax functions are estimated using tax
liabilities before these credits are applied.
Transfers In the Appendix, we provide a description of the various means-tested pro-

grams in the United States, focusing on who quali…es for them and how household’s marital
22
status and number of children a¤ect access.8 We divide these programs into three groups
(i) the Earned Income Tax Credit (EITC); (ii) child-related transfers, that encompass the
Child Tax Credit (CTC), the Child and Dependent Care Tax Credit (CDCTC) and childcare
subsidies; and (iii) the amalgam of programs that provide cash or in-kind transfers that are
routinely identi…ed as the “welfare system", such as the Temporary Assistance to Needy
Families (TANF) and the Supplemental Nutrition Assistance Program (SNAP). We calcu-
late that expenditures in all these programs at all levels amounted to about 2.3% of GDP in
2019.9
S
In computing the transfer functions T RfS (I; k; D); T Rm (I); and T RM (I; k; D); the main
object of this paper, we model tax credits exactly as they appear in the tax code. Following
the discussion in Guner, Kaygusuz and Ventura (2020), the government covers 75% of the
childcare costs for households whose income is below a threshold. We chose the threshold so
that the poorest 5% of children receive the subsidy.
The …nal component is the means-tested transfers. Following Guner, Rauh and Ventura
(2023), we use data from the Survey of Income and Program Participation (SIPP) to estimate
an e¤ective transfer schedule that relates transfers received by di¤erent household types to
their income. The welfare payments include all the main means-tested programs as described
in the Appendix. We assume that these transfers as a function of income take following form
! 0 if I = 0
W (I) = ;
maxf0; ! 1 ! 2 Ig if I > 0
where ! 0 is the transfers for a household with zero income and ! 2 is the bene…ts reduction
rate. Our estimates show that a single female with two children receives about 12% of mean
household income in the economy in terms of welfare transfers (about $12,000 in 2019).
Transfers decline gradually with income and vanish at around 1.1 times mean income for a
single female with two children (about $108,000 in 2019). A single female with two children
and half of mean household income (about $44,000 in 2019) receives about $5,800 per year.
A married couple with two children who has zero income, gets about $8,800. Transfers
decline to zero, as they do for a single mother, at around 1.1 times the mean income.
8
More extended discussions can be found, among others, in Mo¢ tt (2003b) and Guner, Rauh, and Ventura
(2023).
9
In 2019, the U.S. Federal government spent 361 billion dollars for non-medical means-tested transfer
programs for the working-age population. This is about 8% of total federal budget and correspond to about
1.7% of the U.S. GDP. The total spending, federal and state-level, amounted to about 2.3% of the GDP and
is expected to grow as we write. Total spending at all levels is calculated based on information from Rector
and Menon (2018).
23
Figure 4 shows how the total transfers (the sum of these three components) vary by house-
hold income in the benchmark economy. Households without any income receive transfers
in excess of $8,000 per year. The transfers decline sharply for household with positive but
very low income. After that, transfers bounce back to around $8,000 and decline smoothly
with household income and amount to about $1,000 for households with 1.5 times the mean
household income in the economy.
Childcare Costs To determine the requirement of e¢ ciency units for childcare, dM (x; z; t);
and dS (x; t), we use data on total spending (as a fraction of household income) on childcare
and the relation between children’s age and childcare spending (as shown in Figure A4 in
Appendix). In particular, we use data from the Survey of Income and Program Participa-
tion (SIPP), and estimate a relationship between spending in childcare per child and the
average age of children, conditional of the mother’s skill and marital status. Given the price
of unskilled labor services, we recover the e¢ ciency units required at each age in stationary
equilibrium.
10000
8000
6000
Transfers, $
4000
2000
0
0 50000 100000 150000 200000

Household Income (normalized)
Figure 4 - Total Transfers in the Benchmark Economy
Remaining Parameters We select the remaining parameters to match jointly several

targets. (i) We set the Frisch elasticity parameter to 0.2, and given , select the parameters
Bm and Bf to match average market hours per worker by gender. The disutility of work
shocks are speci…ed as L = exp( ) and H = exp(+ ), and 4 is set so as to reproduce the
observed variance of log-hours of married females at age 40. We choose the discount factor to
match capital-to-output ratio (2.9) (ii) We parameterize the distribution of the disutility of
joint market work, (qjz); as a Gamma distribution and infer its parameters to generate the
24
observed female force participation by married females conditional on the husbands’types.
Given (qjz); we determine the loading factors #x (tmin ) so that the model is consistent with
the participation rate of mothers by the age of their youngest child present at home (shown
in Figure A4 in Appendix). (iii) We set the capital share to = 0:343 and the depreciation
k
rate of capital to = 0:055. To select the parameter governing the elasticity of substitution,
, we use standard estimates of this elasticity that suggest a value of 1.5 – see, e.g., Katz
and Murphy (1992). This dictates = 1=3. To calibrate the share parameter , we force the
model to reproduce the aggregate skill premium in the data, de…ned as per-worker earnings
of workers in the skilled category to per-worker earnings of workers in the unskilled category.
For this statistic, we target a value of 1.8. (iv) Finally, we pick the additional proportional
tax, k; on capital so that the model matches corporate tax collections from data, and select
the social security bene…ts, b; for a given tax rate from the US data, to balance the social
security budget.
5 The Benchmark Economy

In Table 1, we show summary statistics on how the model performs regarding targeted and
non-targeted moments. Total transfers in the model are about 2.3% of the GDP, which
(endogenously) matches the data counterpart. The model reproduces the growth in disper-
sion in hourly wages for married individuals by skill, the correlation of wages of married
couples at the start and the end of the life cycle, and married females’participation rates.
Di¤erently from other papers in the literature, the model is in line with skill premia. Among
other factors, this is driven by the fact that rental rates for labor services di¤er by skill as
skilled and unskilled e¢ ciency units are not perfect substitutes in production.
Importantly, the model is in line with the (non-targeted) initial level and growth in house-
hold consumption dispersion over the life cycle (Figure A8). The growth in consumption
inequality is lower than the growth in earnings dispersion for males or for households as
we noted in section 2. One reason for this …nding is that several factors contributing to
dispersion in earnings with age are anticipated as of the start of the life cycle (an aspect
emphasized by Huggett et al., 2011). Furthermore, transfers contribute to total household
income at the bottom of the income distribution and the total household income matters for
consumption inequality. Finally, labor supply adjustments along the intensive and extensive
margin help households smooth idiosyncratic wage shocks.
25
The bottom panel of Table 1 shows earnings inequality measures in the model and the
data for households with heads between ages 25 and 65. The model captures the 90-10 and
90-50 ratios very well, and is able to produce earnings shares of the bottom 10%, 20% and
40% of households, which is critical for the analysis at hand.10 Not surprisingly, taxes and
transfers reduce income inequality non trivially; the 90-10 ratio for after-transfer household
income (after tax), ages 25-65, is of about 6.8 (6.2), while for before taxes and transfers
household income amounts to 7.9.
In sum, the benchmark economy is able to generate observed inequality in the data, it
matches the total spending in transfers as a fraction of GDP when various transfer programs
are modeled as closely as possible to how they operate, and it delivers the level,and the
growth in consumption inequality observed in the data. Altogether, these features suggest
that the model economy is a suitable framework for studying the welfare state.
Life-cycle statistics Our model environment is consistent with a host of observations

over the life cycle. We start by noting that our economy generates the observed growth in
dispersion in hourly wages by skill, gender, and marital status. Figure 5 illustrates this.
We now concentrate on three, interconnected life-cycle statistics. First, we note that our
economy generates the life-cycle pattern of the wage-gender gap, as Figure 6 (left panel)
demonstrates. The model, parameterized to generate the decline in the gender gap by skill
in the early ages of the life cycle, captures quite well the slow opening of the gap for unskilled
workers over the entire life cycle. The model generates the gradual opening of the gap for
skilled workers but leaves a portion unaccounted for towards the end of the working life
cycle. At age 50, skilled females earn 66% on average relative to men in the data, while the
model implies a gender gap of 76%.
10
Inequality measures in the data are from 2010 CPS, with sample restrictions as detailed in the Appendix.
As pointed out by Kuhn and Rios-Rull (2016), there is also signi…cant inequality among single and married
households as well. The model delivers substantial within-group heterogeneity for married and single house-
holds. The 90-10 earnings ratios for married and single households are 5.5. and 6 in the model. The 90-50
ratios are 2.3 and 2.4, respectively.
26
Married Females
Married Males Single Males Single Females
0.40
0.50
0.50
0.40
0.35
0.35
0.40
0.40

0.30
0.30
0.30
0.30
0.25
0.25
0.20
0.20
0.20
0.20
25 30 35 40 45 50 55 25 30 35 40 45 50 55 25 30 35 40 45 50 55
Age Age 25 30 35 40 45 50 55 Age
Age
Unskilled, Data Skilled,Data Unskilled, Data Skilled,Data Unskilled, Data Skilled,Data
Unskilled, Data Skilled,Data
Unskilled, Model Skilled, Model Unskilled, Model Skilled, Model Unskilled, Model Skilled, Model
Unskilled, Model Skilled, Model
Figure 5 - Variance of Log Wages, Model vs. Data, Males (left), Females (right)
Figure 6 (right panel) shows the performance of the model regarding participation rates
of married females as they age. The reader should recall that the economy is parameterized
to reproduce the aggregate levels of participation rates by household type, and their levels
as of age 40. The endogenous forces inside the model – costly children and utility costs
of joint participation that vary with the age of children – lead to the horizontal S-like
pattern of participation rates of married females in the data, as the …gure demonstrates. The
model environment also captures well the initial rise and slow decline of unskilled married
females. Overall, this leads the model economy to reproduce well the aggregate pattern of
participation rates as individuals age.
Overall, as a result of all the forces of our economy operating in tandem, our model implies
an age pattern of dispersion in earnings for married females that is broadly consistent with
observations. Recall from section 2 that dispersion in wages of married females …rst rises,
and unlike the case of men, it ‡attens out as of age 35. As Figure 5 shows, our model
generates the same patterns. Why? Early in the life-cycle, skilled females increase their
skills faster as a group relative to their unskilled counterparts. This, in conjunction with
life-cycle shocks, leads to the overall increase in earnings inequality. In the meantime, some
women gradually return to work –given the gradual reduction in childcare and utility costs of
joint participation as children age –and start increasing their skills by acquiring experience.
Since their skills are lower but accumulate faster, inequality …rst grows but subsequently
starts leveling o¤. Eventually, all di¤erential rates in skill formation become less and less
important as individuals age, and females become more homogenous. The net result is a ‡at
pro…le of earnings dispersion after middle age, as the …gure shows.
27
Children and Childcare Costs What is the quantitative importance of children and
childcare costs? To answer this question, we set all childcare costs to zero, while keeping
all other parameters constant. We …nd that childcare costs matter critically in determining
the levels of participation rates, and how inequality in wages and earnings evolve over the
life-cycle for married females. When childcare costs are set to zero, the participation rate of
unskilled married females is 74.7%, while for skilled, it is 81.4%. The values in the benchmark
model are 68.7% and 77.7%, respectively. The model cannot generate the observed sharp
decline in labor force participation early in the life cycle, demonstrating it is associated
with childcare requirements. Furthermore, without children, the variance of log wages grows
linearly along the life cycle for women, exactly as it does for men (see Figure A9 in the
Appendix).
Gender Wage Gap, Unskilled, Model vs. Data Gender Wage Gap, Skilled, Model vs. Data
Married Female Labor Force Participation, Model vs. Data
1.00
1.00
0.85
0.80
0.90
0.90
Labor Force Particiption

0.75
Gender Gap
Gender Gap
0.80
0.80
0.70
0.65
0.70
0.70
0.60
0.60
0.60
25 30 35 40 45 50 55
25 30 35 40 45 50 55 25 30 35 40 45 50 55 Age
Age Age
Unskilled, data Skilled, data Unskiled, model Skilled, model
Unskilled, model Unskilled, data Skilled, model Skilled, data
Figure 6 - Gender Wage Gap (left); LFP of Marr. Females (right), Model vs. Data
28
Table 1: Model and Data
Aggregates Data Model
Capital Output Ratio 2.9 2.9
Total Transfers (% of GDP) 2.3 2.3
Skill Premium 1.8 1.8
LFP of Married Females (%), 25-54
Unskilled 68.3 68.7
Skilled 77.6 77.7
Total 71.5 72.3
Life-Cycle Inequality
Variance log-wages (Married Males, age 54, S) 0.45 0.45
Variance log-wages (Married Males, age 54, U) 0.34 0.34
Variance log-wages (Married Females, age 54, S) 0.35 0.35
Variance log-wages (Married Females, age 54, U) 0.26 0.26
Variance log-hours (Married Females, age 40) 0.13 0.13
Correlation Between Wages of Spouses (age 25) 0.31 0.31
Correlation Between Wages of Spouses (age 40) 0.34 0.33
Variance log-consumption (Age 54 vs 25) 0.12 0.12
Earnings Inequality (25-64)
90-10 ratio 7.8 7.1
90-50 ratio 2.6 2.5
Share, bottom 10% 1.8 2.1
Note: Entries summarize the performance of the benchmark model in terms of empirical targets and key
aspects of data. The data for aggregate inequality statistics takes into account the same data
restrictions used in the empirical analysis in Section 2.
5.1 How Valuable is the Welfare State?

How much do households value the current transfer scheme? What would be the e¤ects
of abolishing the welfare state? To answer these questions we proceed by fully eliminating
all transfers comprising the welfare state. We balance the budget by adjusting the ‘level’
parameter of the tax function ( ) in a proportional and symmetric way for all households.
Further, as in all subsequent experiments that we conduct, we assume that the rate of
return on capital does not change across steady states. In the Appendix, we also discuss the
elimination of di¤erent transfers one at a time.
Table 2 presents the main …ndings. Hours worked increase across the board, and these
increases are concentrated among the unskilled. The participation rate of married females
29
increases by 6.3% for unskilled women and by 2.0% for skilled ones. All this contributes to
a total increase in labor hours of about 3.0% and an increase in aggregate output of 1.7%.
The average tax rate at mean income falls substantially, by more than four percentage points
across the board; from about 9.2% to 4.8% for a married household with two children, and
from about 7.7% to about 3.3% for a single female with two children.11 When transfers
are eliminated, labor supply increases for low and middle-income households. The increase
in labor supply is partly due to income e¤ects and partly due to the incentives to increase
labor supply for insurance motive. These changes occur despite the removal of programs
that provide incentives for labor supply (e.g., childcare subsidies via the CCDF) or include
provisions that subsidize work (e.g., EITC). For all households, the work incentives also
increase because of lower taxes.
Table 2 also shows large negative e¤ects on aggregate welfare, with a compensating
variation of about -3.2% for all newborn households. Benchmark transfers are substantial
and concentrated at the bottom of the skill distribution. Hence, their elimination leads to
signi…cant welfare losses in a utilitarian sense. Single females bear the brunt of the transfer
elimination. A newborn, single unskilled female experiences a loss of 5.2% on average, while
an equivalent single skilled female faces a loss of 1.2%. Nonetheless, since tax rates fall
substantially, a majority of adults bene…t from the elimination of the welfare state –about
60.7% of households bene…t.
Overall, these …ndings highlight and anticipate trade-o¤s associated with reforming the
welfare state. The welfare state targets transfers to low and middle-income households.
As a result, while they depress participation, hours, and output, they are highly valuable
for some households and translate into substantial losses for all newborns associated with
its elimination – even when tax rates are sharply reduced across the board. These losses
mask gains for many agents, resulting in a signi…cant majority of newborns in favor of this
hypothetical move. The signi…cant majority in favor of elimination of the system (60.7%)
illustrates the trade-o¤s involved in an economy with substantial heterogeneity like ours.
11
To balance the budget in this exercise, we multiply values in Table A6 in Appendix by 1.048 (recall
that 1 is the tax rate at the mean income).
30
Table 2: Eliminating All Transfers
Aggregates Welfare
(% changes relative to benchmark) (Newborns, %)
Output 1.7 Single F
Aggregate Hours 3.0 Unskilled -5.2
Hours per worker (All Females) 3.2 Skilled -1.2
Hours per worker (All Males) 1.9 Married
Unskilled, Unskilled -0.1
Participation Married Females Unskilled (f), Skilled (m) 0.4
Unskilled 6.3 Skilled, Skilled 1.7
Skilled 2.0 Skilled (f), Unskilled (m) 0.6
Total 4.5 All
All Newborns -3.2
Winning Households 60.7
Note: Entries in the left panel show the e¤ects (percentage changes) across steady states on selected aggregate
Entries in the right panel show the corresponding welfare e¤ects (consumption compensation)
for newborn households.
6 Rethinking the Welfare State

We now conduct several quantitative experiments in which we provide answers to the ques-
tions that motivate the paper. In all experiments, the rate of return of capital is constant
across steady states – but rental prices for labor services change in order to be consistent
with equilibrium conditions. We …rst consider replacing the current transfer scheme with a
Universal Basic Income scheme (UBI) and then with a Negative Income Tax (NIT) across
steady state equilibria. We then discuss the e¤ects of these reform cases when transitional
dynamics are taken into account. In the benchmark economy, tax revenue is used to …nance
all the transfers and government spending G. When we eliminate existing transfers, we
assume that the government adjust taxes to cover G and transfers associated with UBI or
NIT.
6.1 A Universal Basic Income

In our …rst experiment, each household receives a transfer per household member (including
children) in all dates and states. The current welfare state is abolished while the tax system
is unchanged. We dub this experiment a Universal Basic Income scheme (UBI). Speci…cally,
we search across steady states for the level of the UBI transfer that maximizes the ex-ante
31
welfare of all newborns. We balance the budget by adjusting the ‘level’parameter of the tax
function ( ) proportionally.
Our …ndings are presented in Tables 3 and 4. We …nd that a per-person transfer of about
3.2% of mean household income maximizes the welfare of newborns. This corresponds to
about $3,100 per person in 2019 dollars ($12,400 for a married household with two children
at home). To balance the budget, tax rates need to substantially increase; for a married
household with two children at mean income, the average rate increases to 14.4% from
9.2% in the benchmark.12 This occurs as at the welfare-maximizing level, the aggregate
expenditure on transfers increases in a signi…cant way relative to the benchmark; from 2.3%
in the benchmark case to 5.9%. The UBI transfers, coupled with higher taxes, depress
hours, participation and output across steady states. Total hours and output decline by
0.9% and participation rates of unskilled and skilled married females decline by 4.4% and
1.9%, respectively. Since hours worked for those away from the margin of indi¤erence do
not change much relative to the benchmark, per-worker hours for females increase, as Table
3 shows.
Table 4 illustrates the welfare consequences of the UBI policy across steady states. Even
at the best policy, ex-ante welfare for all newborns declines, with a compensating variation
of -1.3%. But a majority of newborns, 53.2%, support a UBI program. As it was the case
with eliminating the current welfare system, lifetime-poor households su¤er under the UBI,
since it does not fully replace the transfers they were receiving in the benchmark economy
while they see their taxes go up. This contributes to an overall welfare loss. Unskilled single
females experience a welfare loss at birth of about 2.5%, and skilled ones a loss of about 0.7%.
On the other hand, unskilled married households are strong winners as Table 4 demonstrates,
with a welfare gain of about 2.1% for a married couple with two unskilled adult members.
This re‡ects that some low-to-middle income households, who did not receive much in terms
of transfers in the benchmark economy, now get a generous UBI transfer contributing to
generating majority support.
Transitional Dynamics What is the role of transitional dynamics in our welfare …nd-
ings? To address this question, we compute the transitions associated to the non-anticipated
introduction of the welfare-maximizing transfer discussed before. The budget is balanced
by adjusting the level parameter of the tax function in each period of the transition across
12
The values in Table A6 are multiplied by 0.9433 to balance the government’s budget.
32
steady states.
Table 4 shows that the welfare losses are lower for newborn households at the start of
the transition than in the steady state – 0.5% vs 1.3%. As the economy gradually shrinks
along the transition, taxes to …nance the UBI transfer increase, and therefore, welfare losses
are larger when steady states are compared. Nonetheless, there is a majority of newborns
who lose from the implementation of a UBI; the fraction of newborns who support the policy
drops from 53.2% to 44.1%. If we only count all households alive at the start of the transition,
there is still a majority against the implementation.
A UBI on top of the Welfare State? It is worth noting that the welfare-maximizing
transfer level is substantially below the magnitudes advocated in policy discussions. For
instance, there are proposal of a UBI transfer of $1,000 per month per adult, which is a
much higher transfer than what we …nd as optimal. Indeed, we …nd that with transfer
levels higher than the optimal one, welfare losses non-trivially increase and popular support
dissipates quickly,
A natural next question is: what if the UBI transfer is given on top of the existing welfare
state? We …nd that a relatively small transfer of about 0.5% of mean household income
maximizes ex-ante welfare. There is a welfare gain of 0.1%, but only 48.9% of households
are in favor of such a program. Larger transfers lead to welfare losses. For instance, if the
transfer is 1% of the mean household income per person, an ex-ante welfare loss emerges and
52.5% of adults oppose it. If instead, we impose the transfer that maximizes welfare under
the UBI reform (a transfer of 3.2% of mean income) on top of the existing programs, output
losses are more signi…cant, and ex-ante welfare and popular support decline much further.
These …ndings follow from the fact that the welfare system is resilient, as its elimination
leads to large welfare losses. Likewise, a common transfer to all is quite expensive and requires
non-trivial tax hikes that depress welfare. We conclude from these …ndings that only small
transfers on top of the existing welfare state –far from those advocated by proponents of a
UBI –can improve marginally welfare but without majority support.
Summary Overall, our …ndings indicate that a UBI policy reform is hard to justify on
ex-ante welfare grounds as a replacement for the current welfare state. Yet, it is supported by
a majority despite its macroeconomic magnitude and the additional tax revenue it requires.
A UBI policy on top of the current welfare state is not a good idea; only marginal welfare
33
gains emerge for a small transfer and there is clear majority against the move.
6.2 A Negative Income Tax

We now evaluate a more drastic reform that eliminates the current welfare state and the
progressive income taxation. Speci…cally, we introduce a proportional income tax combined
with a transfer for all, adult and children. Following Friedman (1962) and the literature that
followed, we dub this linear income tax a Negative Income Tax system, or NIT for short.
We again search for the welfare-maximizing per household member transfer and balance the
budget by adjusting the proportional tax rate that applies to all households.
Table 3 shows the e¤ects on aggregates. The transfer at the welfare-maximizing level
is about 4.8% of mean household income, or $4,700 in 2019 dollars ($18,800 for a married
couple with two children). Thus, the welfare-maximizing NIT transfer is signi…cantly more
generous than the best one in the UBI case, and involves a drastic increase in resources
devoted to redistribution –about 8.8% of output. The proportional tax rate that supports
the welfare-maximizing NIT is 19.8%. Thus, tax rates are higher for most households – a
married household with two children at around mean income faces an average tax rate of
about 9.2% in the benchmark economy – but marginal rates are lower for those with top
incomes. For instance, the marginal rate for a married household at three times mean income
amounts to 21.4% in the benchmark case.
Overall, in net terms, the NIT reform leads to depressing e¤ects on hours worked and
output, as Table 3 demonstrates. Indeed, given its generosity at the welfare-maximizing
case, its e¤ects on total hours worked and participation rates are stronger than in the UBI
case. But output declines less with the NIT than it does with the UBI (0.9% vs. 0.6%) since
a proportional tax encourages higher labor supply from more productive households.
Welfare Table 4 shows that an NIT generates ex-ante welfare gains of about 0.2% of
consumption, which are accompanied by substantial majority support for the reform among
newborns – more than two thirds of newborns support the reform at birth. There are of
course winner and losers. Married households enjoy substantial welfare gains, which is most
signi…cant among couples with two unskilled partners. On the other hand, single female
households as a group experience ex-ante losses. The losses are more signi…cant among
unskilled and those with children.
In the Appendix (see Figure A10) we show how aggregate output, ex-ante aggregate
34
steady state welfare, and majority support changes for di¤erent levels of as the NIT transfers.
When the transfer equals zero, i.e., when the tax system is simply a proportional tax with
no transfers whatsoever, output is about 3.2% higher than in the benchmark case. But there
is a substantial welfare cost. As transfers increase, tax rates, welfare and popular support
increase as well, but output declines. Altogether, as the lump-sum transfer increases, both
welfare and support for the reform …rst sharply increase and then decline when the decline
in output dominates.
If we consider transitional dynamics, the welfare gains associated to the welfare-maximizing
transfer among newborns at the start the transition are higher than under the comparisons
across steady states (0.4% vs 0.2%). A majority of newborn households support the policy
shift (55.1%). This majority is even higher (59.4%) if we consider all households alive at the
start of the transition.
Table 3: Aggregate and Inequality Findings

(changes relative to benchmark)
UBI NIT NIT (2)
Maximum Maximum Maximum
Welfare Welfare Welfare
Output -0.9 -0.6 -0.4
Aggregate Hours -0.9 -1.2 -1.0
Hours per worker (All Females) 0.8 0.5 -0.0
Hours per worker (All Males) -0.5 -0.5 -0.2
Participation Married Females:
Unskilled -4.4 -6.0 -4.2
Skilled -1.9 -2.3 -1.5
Total -3.4 -4.4 -3.1
Proportional Tax Rate (%) - 19.8 19.8

Transfer (% Household Income) 3.2 4.8 7.0, 4.1
Transfers (% Output ) 5.9 8.8 8.7
Inequality
4 Var Log-Household Income -0.027 0.01 0.009
4 Var Log-Household Income -0.006 0.029 -0.014
(after taxes and transfers)
Note: Entries in the top panel show e¤ects (percentage changes) across steady states on selected variables
driven by the di¤erent quantitative experiments. Entries in the bottom panel are changes in the variance
of log-household income before and after tax and transfers, relative to the benchmark economy.
35
Comparison with a Proportional Income Tax Since the NIT has two elements,
i.e., the replacement of all transfers with a common payment to individuals and a move to a
proportional tax system, it is informative to study what happens with a simple proportional
tax that leaves the welfare state in place. We …nd that in this case, aggregate hours and
output increase by about 1.6% and 1.4%, respectively, requiring a supporting tax rate of
10.8%. In terms of welfare, we …nd e¤ectively no gains or losses on an ex-ante basis. Again
there are sharp di¤erences between winners and losers. Our results show that skilled married
couples have a welfare gain of 1.1%, whereas unskilled single females experience a loss of
about 0.4%. A strong majority of newborn households, about 62.3% of them, are against it.
Hence generous transfers, made possible by a proportional income tax, is key for the success
of NIT.
Di¤erentiated Transfers Since welfare gains from an NIT reform are unevenly dis-
tributed between married and single households and relatively small in the aggregate, we
consider an NIT regime with transfers di¤erentiated by the marital status of adults but with
a common tax rate. We refer to this case as NIT(2). Speci…cally, we search for a transfer
and ratio of transfers to individuals in married households relative to single households that
maximize ex-ante welfare and preserve majority support.
The welfare-maximizing transfer per person in single households is about 7% of the
mean household income, while about 4.1% in married households (about $6,860 and about
$4,000 in 2019 dollars). The tax rate that supports this arrangement is about the same
as in the baseline NIT exercise (19.8%). Output and aggregate hours decline by 0.4% and
1.0% relative to the benchmark economy. This reform e¤ectively means that a single female
with 2 children, under an income level of one-half mean household income, would receive a
net transfer after taxes of about 11.1% of mean household income (about $10,850 in 2019
dollars). The net transfer for a married couple with the same income and two children would
be about 6.50% of mean household income (about $6,350 in 2019 dollars).
An NIT reform of this type leads to ex-ante welfare gain of about 0.7%, with 51.4% of
adults supporting the reform in the welfare-maximizing scenario. Clearly, gains are larger
than in the undi¤erentiated NIT. As earlier, if we consider transitions across steady states,
welfare gains for newborn households are larger, and a substantial majority supports the
policy move.
36
Inequality The alternative tax-transfer schemes have ambiguous e¤ects on inequality,
as Table 3 shows. The variances of log-household income, before and after taxes and transfers,
decline with a shift to a UBI. Meanwhile, both measures increase with the NIT case, and
move in opposite directions in the case of an NIT with di¤erentiated transfers for marital
status.
In understanding these results, it is key to bear in mind that several forces are at play
here. First, the removal of transfers in the benchmark and the replacement for a smaller UBI
transfer leads to increases in labor supply for some groups (e.g., low-skilled single females),
contributing to a reduction in inequality in the UBI case. Second, an NIT reform involves
a larger transfer alongside the full elimination of increasing marginal rates on household
income. The result is a net increase in household income inequality. The same forces, with
di¤erentiated transfers dictate larger increases in inequality before taxes and transfers in the
second NIT experiment. Overall, the e¤ects on inequality of drastically di¤erent alternatives
appear of second order.13
Summary Quantitatively, welfare gains under an NIT regime are substantially larger
than the (negative) gains under a standard UBI scheme, and even larger gains can be ob-
tained when NIT transfers di¤erentiated by marital status. Taking into account transitional
dynamics makes these welfare …ndings even stronger. Overall, what accounts for the relative
success of a Negative Income Tax in terms of ex-ante welfare and majority support? The up-
shot is that a larger degree of redistribution is feasible given the smaller tax distortions under
an NIT regime; i.e., with the elimination of increasing marginal tax rates and lower taxes on
secondary earners. As tax distortions are reduced with a proportional tax, compared to the
UBI case, the size of the aggregate economy is larger and collecting the tax revenues that
are necessary to …nance transfers becomes easier. The net result is that a higher transfer
level becomes feasible under an NIT relative to a UBI scheme. Put di¤erently, a drastic tax
reform that reduces marginal tax rates at top incomes while making them common across
all earners, makes more extensive redistribution possible.
13
The changes in the log-variances are quite small relative to the changes in variances due to taxes and
transfers. For instance, while a baseline NIT increases the pre-tax and transfer inequality by about 1 log
point, taxes and transfers reduce the variance by about 14.7 points.
37
7 Findings in Perspective
In this section, we place our results in perspective, with a focus on the e¤ects and conse-
quences of an NIT. We study an NIT reform in an economy with a lower level of inequality
than our benchmark. We then evaluate an NIT reform when life-cycle facts are parameter-
ized under a cohort instead of time e¤ects. We also assess the use of progressive taxation to
…nance transfers and the e¤ects of changes in the parameterization of preferences.
7.1 Rethinking the Welfare State When Inequality is Lower

The U.S. economy has changed in critical ways relative to what it was in the recent past. In
particular, there have been drastic changes in the demographic composition of U.S. house-
holds and in dimensions of inequality. For instance, only about 19% of females were skilled
under our de…nition in 1980, while this …gure more than doubled to nearly 39% in 2008,
our baseline year for demographics. Substantial changes in marital sorting accompanied
these changes; about 14% of married households were of the skilled-skilled category in 1980,
while the corresponding …gure in our parameterization is nearly 27%. Meanwhile, the skill
premium was 1.4 under our assumptions in 1980, and it increased to 1.8 in our benchmark
parameterization.
Since these changes can a¤ect the importance of the redistributive component of the NIT
vis-a-vis its e¤ects on tax distortions, they could be of importance for our conclusions. To
what extent these underlying di¤erences matter for the implications of an NIT? To address
this question, we parameterize our economy to demographic and endowment targets of the
past (circa 1980). We impose the demographics prevailing in 1980, use corresponding wage
inequality over the life cycle to parameterize endowments, and force our economy to be
consistent with the skill premium in 1980. We refer to this case as the 1980 economy. We
then search for the welfare-maximizing NIT as in our baseline exercises.14
We conduct our exercises in two levels. In our …rst case, we do not impose the partici-
pation rates of 1980 to parameterize the 1980 economy, while we do so in our second case.
We summarize our results in Table 5. We …nd that the welfare-maximizing NIT does not
lead to a contraction in output in the long run. Furthermore, the resulting welfare gains are
larger than in our baseline case (0.2% vs 0.7-0.8%), with more substantial majority support,
under an NIT transfer that is smaller – about 3% of household income. In short, these
14
We describe the 1980 data in detail in the Appendix.
38
…ndings follow from the factors leading to lower underlying inequality among households.
These factors determine that an NIT reform requires a lower transfer to maximize ex-ante
welfare and a concomitant lower tax rate than in the baseline case. Hence, the detrimental
e¤ects on aggregates from taxes and transfers are of smaller magnitude, resulting in turn in
the expansion (not contraction) of output and hours that Table 5 shows.
We conclude from these …ndings that an NIT reform in an economy with the characteris-
tics of the United States circa 1980 is more appealing on several fronts, regardless of whether
we consider changes in participation rates or not (case I or case II). Di¤erent factors leading
to lower levels of inequality result in an environment more favorable to the implementation
of an NIT reform.
Table 4: Welfare E¤ects

(Consumption Compensation, Newborns, %)
UBI NIT NIT (2)
Welfare Welfare Welfare
Single F
Unskilled -2.5 -2.0 0.5
Skilled -0.7 -0.7 0.3
No Child -0.6 -0.6 0.3
Early Childbearing -2.2 -1.7 0.3
Late Childbearing -0.5 -0.4 0.2
Married
Unskilled, Unskilled 2.1 2.8 0.2
Unskilled (f), Skilled (m) 0.2 0.3 -0.1
Skilled, Skilled 0.0 0.6 -0.3
Skilled (f), Unskilled (m) 0.2 0.4 -0.1
No Child 0.1 0.1 -0.2
Early Childbearing 1.4 2.3 -0.1
Late Childbearing 1.0 1.7 -0.0
All Newborns -1.3 0.2 0.7

Winning Households 53.2 68.2 51.4
Including Transitional Dynamics

All Newborns -0.5 0.4 1.0
Wining Households 44.1 55.1 56.8
Wining Households (All Alive) 46.8 59.4 75.3
Note: Entries show the welfare e¤ects (consumption compensation) driven by the reform of the welfare state.
The top panel report welfare gains across steady states with a constant interest rate across steady states.
The calculations in the panel below take into transitional dynamics between steady states.
39
7.2 Cohort E¤ects in Data
As described in sections 3.1 and 4, our analysis relies on a parameterization in which life-
cycle facts were characterized controlling by time e¤ects. In the Appendix, we describe our
main empirical …ndings when we extract age pro…les controlling instead by cohort e¤ects.
Under cohort e¤ects we …nd steeper pro…les of hourly wages as well as steeper pro…les of
wage inequality relative to the time e¤ects case. These …ndings are in line with the literature.
What are the implications of assuming cohort e¤ects in our parameterization for our
results? Table 5 summarizes our …ndings. We …nd a welfare-maximizing NIT transfer of
around 4.7% of mean household income, roughly at the same level of the baseline case under
time e¤ects. Welfare gains are higher, at about 0.8% for all newborn households. Aggregate
output and hours expand across steady by 0.7% and 0.4%, respectively, unlike the baseline
case under time e¤ects.
Qualitatively, these results are unsurprising. Steeper inequality pro…les lead to higher
variances of persistent shocks relative to the time e¤ect case, and therefore, there is a larger
role of transfers for all provided by an NIT transfer. Likewise, steeper wage pro…les lead to
potentially larger gains related to the replacement of increasing marginal rates of household
taxation. Hence, larger gains welfare gains are available. We conclude from these …ndings
that parameterizing our economy under a cohort e¤ect view of the life-cycle data makes a
basic NIT reform more desirable.
7.3 Funding Transfers with Higher Progressivity

In our main exercises, common transfers are funded by taxing household income in the least
distorting ways possible; either by shifting the tax function (UBI) or by using a proportional
income tax (NIT). What if transfers are …nanced by changing the progressivity of the tax
function? Can this improve upon the UBI or NIT schemes analyzed earlier?
To address this question, we proceed as follows. We eliminate existing transfers, and
search for the level of tax progressivity (curvature) of the income tax function and a common
transfer to all individuals that are consistent with budget balance and maximize ex-ante
welfare. Since the curvature of tax functions di¤ers (married, single, etc.), we search for the
common factor that increases or reduces the level of progressivity in all cases. Note here
that increasing the curvature of the tax function implies a rotation in the tax function; it
increases average rates at the top, but reduces average tax rates at the bottom.
We …nd a welfare-maximizing transfer relatively small, of only about 1.25% of mean
40
household income, and that this transfer level implies a non-trivial reduction in the curvature
of the tax functions – about 60% reduction across the board. This determines an ex-ante
welfare loss of about -1.8%. Put di¤erently, the welfare-maximizing transfer-progressivity
combination does not dominate the status quo and leads to a worse outcome than our basic
UBI.
To understand these …ndings, it is important to note that increases in progressivity from
the benchmark levels, as they lead to steeper marginal rates, lead to non-trivial declines in
economic aggregates and can only increase revenue marginally in the long run. Guner et al
(2016) make explicitly this point in the context of life-cycle model with heterogeneity. Using
revenues to provide for transfers to all make the problem of obtaining additional revenues
even worse. We conclude from these …ndings that a shift to common transfers to all hinge
critically on imposing taxes that distort decisions little, either by shifting taxes for all (UBI)
or by drastically changing the nature of income taxation (NIT).
Table 5: Findings in Perspective (% changes relative to benchmark)

Baseline 1980 1980 Cohort CRRA & Scale
Findings (I) (II) E¤ects Economies
Output -0.6 0.7 0.6 0.7 -3.8

Aggregate Hours -1.2 0.9 0.8 0.4 -5.8
Unskilled -6.0 -0.1 -0.3 -1.1 -18.3
Skilled -2.3 0.3 -0.2 1.4 -8,1
Total -4.4 0.1 -0.3 0.0 -13.9
Welfare
All Newborns (%) 0.2 0.8 0.7 0.8 0.8
Winning Households 68.2 80.0 81.0 67.5 49.5
Proportional Tax Rate (%) 19.8 14.8 14.8 19.6 30.8
Transfer (% Household Income) 4.8 3.0 3.0 4.7 8.2
Transfers (% Output ) 8.8 5.9 5.9 8.6 15.6
Note: Entries in the top panel show e¤ects of a welfare-maximizing NIT experiment. The second column
shows the baseline …ndings for comparison. The third column displays the results when shocks and
demographics are calibrated to data circa 1980, but married female participation rates are from the
benchmark. The fourth column shows the corresponding results when shocks, demographics and
participation rates are calibrated to data from circa 1980. The …fth column shows the results when the
benchmark economy is calibrated under a cohort e¤ects-view of the data. The …nal column shows
results when preferences allow for a risk aversion coe¢ cient exceeding 1 and scales economies in
household consumption.
41
7.4 CRRA preferences and Economies of Scale
We have conducted all of our analysis using preferences in which preferences for consumption
are logarithmic, with a coe¢ cient of relative risk aversion of 1. Likewise, our analysis did
not include scale economies in consumption as scale economies are not relevant with log-
preferences, but scale economies could become relevant with CRRA preferences. What are
the implications for our conclusions regarding an NIT reform if CRRA preferences and scale
economies are considered?
To answer this question, we set the relative risk aversion to a higher level (1.5) and impose
a commonly used adjustment for scale economies, equal to the square root of the number of
people in the household. We then parameterize the model economy again in this scenario
and …nd the welfare-maximizing NIT arrangement. Our …ndings are summarized in Table 5.
In this scenario, an NIT experiment leads to much larger transfers and associated tax rates,
and to substantial declines in output and labor supply aggregates. The welfare-maximizing
transfers to all reach 8.2% with a tax rate of about 31%. Welfare gains are of about 0.8%,
but without a majority of households in support of the NIT reform.15
In interpreting these results, it is important to note that this exercise is a major departure
in our analysis and should be taken with caution. Our benchmark parameterization is consis-
tent with a host of facts –including the growth of consumption dispersion over the life cycle
–and in macroeconomic terms, it is consistent with balanced growth. This alternative case
is not consistent with balanced growth, with income e¤ects dominating substitution e¤ects.
This implies that increasing wages for all would reduce labor supply and that increasing tax
rates would increase it. We would then expect that much (higher) transfers emerge in the
welfare-maximizing case, with concomitantly large depressing e¤ects on aggregates as our
results show.
8 Concluding Remarks
Three main points emerge from our analysis so far. First, it is hard to improve upon the
current structure of the welfare state via simple transfer schemes. Transfers to poorer house-
holds are highly valued, and thus, any reform to the current system needs to confront the
15
If we impose a CRRA coe¢ cient of 1.5 without any scale economies in consumption, calibrating the
economy to data, the results are more moderate but similar. Output and hours worked decline by 3.1 and
4.9% across steady states, respectively, under a transfer of about 7.75% of mean household income. Welfare
gains amount to 0.4%.
42
fact that non-trivial resources accrue to poorer households. As a result, simpler schemes
that maximize ex-ante welfare relative to the status quo require drastic changes in taxation.
Second, a UBI scheme is generically not a good idea and is dominated by an NIT. Why?
Considerable resources need to be transferred to poorer households for their welfare not to
fall. And since transfers would accrue to all individuals, taxes need to increase substantially,
leading to ex-ante welfare losses. A generous UBI transfer imposed on top of the current
welfare state, or a UBI transfer …nanced via increases in tax progressivity, do not change this
conclusion. It follows that in our economy, a UBI scheme, as proposed in popular discussions,
is not a good idea.
Lastly, NIT arrangements generate ex-ante welfare gains and lead to popular support
due to the associated reduction or elimination of pre-existing distortions (i.e., increasing
marginal tax rates, household rather than individual income-tax base) and the concomitant
increase in output and revenues. Our …ndings hold when transitional dynamics are taken
into account and when life-cycle data is interpreted via cohort e¤ects instead of time e¤ects.
Interestingly, we also …nd that the desirability of an NIT reform is higher for an economy
with the characteristics of the U.S. economy in the past (circa 1980).
We end this paper with three comments. First, the administrative costs of running a
welfare state can be large. Isaacs (2008), for example, calculates that the cost of running
Food Stamps, Housing Subsidies and the TANF programs are as high as 15 cent per each
dollar bene…t issued. Our analysis abstracts from such administrative costs, and hence might
underestimate the potential bene…ts of moving to a simpler system like the NIT. Second, a
variant to the NIT system could use a consumption tax instead of a ‡at-rate income tax.
As a consumption tax does not distort capital formation, this implementation could lead to
larger gains in output, labor supply and welfare than we found in our analysis. Finally, it
might be important to evaluate reforms to the welfare state when health and health-related
transfers are taken into account. In this vein, we consider that a model of endogenous health
where both medical and nonmedical means-tested transfers are modeled and non-medical
income support can a¤ect investments in health, is a promising line of future research.16 We
leave this and other issues for future research.
16
Mahler and Yum (2022) for instance, build a model of endogenous health where changes in healthy
behavior are subject to adjustment costs.
43
References
[1] Aguiar, Mark and Erik Hurst (2013): "Deconstructing Life Cycle Expenditure," Journal
of Political Economy, 121, 437-492.
[2] Aiyagari, S. Rao, Greenwood, Jeremy and Nezih Guner (2000): “On the State of the
Union,”Journal of Political Economy, 108, 213–44.
[3] Attanasio, Orazio P., Banks, James, Meghir, Costas, and Guglielmo Weber (1999):
"Humps and Bumps in Lifetime Consumption," Journal of Business and Economic
Statistics, 17, 22-35.
[4] Attanasio, Orazio P., Hamish Low, and Virginia Sánchez-Marcos (2005): “Female Labor
Supply as Insurance Against Idiosyncratic Risk,” Journal of the European Economic
Association, 3, 755–64.
[5] Badel, Alejandro, and Mark Huggett and Wenlan Luo (2020): “Taxing Top Earners: A
Human Capital Perspective,”The Economic Journal, 130, 1200–1225.
[6] Bayer, Christian and Moritz Kuhn (2020): "Which Ladder to Climb? Decomposing
Life-Cycle Wage Dynamics," Working Paper, University of Bonn.
[7] Bitler, Marianne, Jonathan Gelbach, Hilary Hoynes, and Madeline Zavodny (2004):
“The Impact of Welfare Reform on Marriage and Divorce,”Demography 41, 213–36.
[8] Blundell, Richard, Luigi Pistaferri and Itay Saporta-Eksten (2016): "Consumption In-
equality and Family Labor Supply," American Economic Review, 106, 387-435.
[9] Blundell, Richard, Monica Costa Dias, Costas Meghir, and Jonathan Shaw (2016):
“Female Labor Supply, Human Capital and Welfare Reform,”Econometrica, 84, 1705–
1753.
[10] Boar, Corina and Virgiliu Midrigan (2022): "E¢ cient Redistribution," Journal of Mon-
etary Economics, 131, 78-91
[11] Margherita Borella, Mariacristina De Nardi, and Fang Yang (2023): "Are Marriage-
Related Taxes and Social Security Bene…ts Holding Back Female Labour Supply?" Re-
view of Economic Studies, 90, 102–131.
[12] Conesa, Juan Carlos, Li, Bo and Qian Li (2023): "A quantitative evaluation of Universal
Basic Income," Journal of Public Economics 223, 104881.
[13] Daruich, Diego and Raquel Fernandez (2023): "Universal Basic Income: A Dynamic
Assessment." NBER Working Paper 27351.
[14] Eckstein, Zvi, Keane, Michael. and Osnat Lifshitz (2019): "Career and Family Decisions:
Cohorts Born 1935–1975," Econometrica, 87, 217-253
44
[15] Ferriere, Axelle, Philipp Grubener, Gaston Navarro, and Oliko Vardishvili (2023): "On
the Optimal Design of Transfers and Income-Tax Progressivity," Journal of Political
Economy Macroeconomics 1(2): 276-333.
[16] Friedman, Milton (1962): Capitalism and Freedom. University of Chicago Press,
Chicago, London.
[17] Greenwood, Jeremy, Nezih Guner, Georgi Kocharkov, and Cezar Santos (2016): "Tech-
nology and the Changing Family: A Uni…ed Model of Marriage, Divorce, Educational
Attainment, and Married Female Labor-Force Participation," American Economic Jour-
nal: Macroeconomics, 8, 1-41.
[18] Goldin, Claudia (2021): Career and Family: Women´ s Century-Long Journey Towards
Equality. Princeton University Press: Princeton, NJ.
[19] Guner, Nezih, Remzi Kaygusuz, and Gustavo Ventura (2012): “Taxation and Household
Labour Supply," Review of Economic Studies, 79, 1113-1149.
[20] Guner, Nezih, Remzi Kaygusuz, and Gustavo Ventura (2020): “Child-Related Transfers,
Household Labor Supply and Welfare," Review of Economic Studies 87, 2290–2321.
[21] Guner, Nezih, Christopher Rauh, and Gustavo Ventura (2023): “Means-Tested Trans-
fers in the US: Facts and Parametric Estimates,”Mimeo.
[22] Guner, Nezih, Lopez-Daneri, Martin and Gustavo Ventura (2016): "Heterogeneity and
Government Revenues: Higher Taxes at the Top?" Journal of Monetary Economics, 80,
69-85.
[23] Guvenen, Fatih (2009): “An Empirical Investigation of Labor Income Processes,” Re-
view of Economic Dynamics, 12, 58-79.
[24] Heathcote, Jonathan, Storesletten, Kjetil, and Gianluca Violante (2005): "Two Views
of Inequality over the Life Cycle," Journal of the European Economic Association, 3,
765-775
[25] Heathcote, Jonathan, Storesletten, Kjetil and Gianluca Violante (2010): “The Macro-
economic Implications of Rising Wage Inequality in the United States,” Journal of
Political Economy 118, 681-72.
[26] Heathcote, Jonathan, Storesletten, Kjetil and Gianluca Violante (2017): "Optimal Tax
Progressivity: An Analytical Framework," Quarterly Journal of Economics,132, 1693–
1754.
[27] Holter, Hans A., Dirk Krueger, and Serhiy Stepanchuk (2019): "How do Tax Progres-
sivity and Household Heterogeneity A¤ect La¤er Curves?" Quantitative Economics,
10,1317-1356.
[28] Hosseini, Roozbeh and Kopecky, Karen A. and Zhao, Kai (2021): "How Important Is
Health Inequality for Lifetime Earnings Inequality?" FRB Atlanta Working Paper No.
2021-1.
45
[29] Hoynes, Hilary and Jesse Rothstein (2019): "Universal Basic Income in the US and
Advanced Countries," Annual Review of Economics, 11, 929–958.
[30] Huggett, Mark, Gustavo Ventura, and Amir Yaron (2011): "Sources of Lifetime In-
equality," American Economic Review, 101, 2923–54.
[31] Isaacs, Julia (2008): The Costs of Bene…t Delivery in the Food Stamp Program. United
States Department of Agriculture, Contractor and Cooperator Report No. 39
[32] Jaimovich, Nir, Saporta-Eksten, Itay, Setty, O¤er and Yaniv Yedid-Levi (2021): "Uni-
versal Basic Income: Inspecting the mechanisms." Working Paper
[33] Katz, Lawrence and Kevin Murphy (1992): “Changes in Relative Wages, 1963-1987:
Supply and Demand Factors," The Quarterly Journal of Economics, 107, 35-78.
[34] Kaygusuz, Remzi (2010): “Taxes and Female Labor Supply,” Review of Economic Dy-
namics, 13, 725-741.
[35] Kaygusuz, Remzi (2015): "Social Security and Two-Earner Households," Journal of
Economic Dynamics and Control, 59,163–178.
[36] Kearney, Melissa S. (2004): "Is There an E¤ect of Incremental Welfare Bene…ts on
Fertility Behavior? A Look at the Family Cap," Journal of Human Resources, 39,
295-325.
[37] Kindermann, Fabian and Dirk Krueger (2020): “High Marginal Tax Rates on the Top
1%? Lessons from a Life Cycle Model with Idiosyncratic Income Risk," American
Economic Journal: Macroeconomics, 14, 319-366.
[38] Kuhn, M. and J. V. Rios-Rull (2016): “2013 Update on the US Earnings, Income, and
Wealth Distributional Facts: A View from Macroeconomics," Federal Reserve Bank of
Minneapolis Quarterly Review 37(1), 2-73.
[39] Lopez-Daneri, Martin (2016): "NIT Picking: The Macroeconomic E¤ects of a Negative
Income Tax," Journal of Economic Dynamics and Control, 68, 1–16.
[40] Low, Hamish, Meghir, Costas, Pistaferri, Luigi and Alessandra Voena (2022): "Mar-
riage, Labor Supply and the Dynamics of the Social Safety Net," NBER Working Paper
No. 24356.
[41] Luduvice, Andre Victor Doherty (2019): "The Macroeconomic E¤ects of Universal Basic
Income Programs," Working Paper, Federal Reserve Bank of Cleveland.
[42] Mahler, Lukas and Minchul Yum (2022): "Lifestyle Behaviors and Wealth-Health Gaps
in Germany," Working Paper, University of Southampton.
[43] Mo¢ tt, Robert A. (2003); "The Negative Income Tax and the Evolution of U.S. Welfare
Policy," Journal of Economic Perspectives, 17, 119-140.
46
[44] Mo¢ tt, Robert J., Phelan, Brian J. and Anne E. Winkler (2020): "Welfare Rules,
Incentives, and Family Structure," Journal of Human Resources, 55, 1-42
[45] OECD. (2019): Social Expenditure Update 2019: Public Social Spending is High in
Many OECD Countries, OECD Publishing, Paris.
[46] Ortigueira, Salvador and Nawid Siassi (2013): "How Important is Intra-Household Risk
Sharing for Savings and Labor Supply?" Journal of Monetary Economics, 60, 650 666.
[47] Ortigueira, Salvador and Nawid Siassi (2022): "The U.S. Tax-Transfer System and
Low-income Households: Savings, Labor supply, and Household Formation," Review of
Economic Dynamics, 44, 184-210.
[48] Primiceri, Giorgio and Thijs van Rens (2009): "Heterogeneous Life-cycle Pro…les, In-
come Risk and Consumption Inequality," Journal of Monetary Economics, 56, 20-39.
[49] Rauh, Christopher and Marcelo R. Santos (2022): "How do Transfers and Universal
Basic Income Impact the Labor Market and Inequality?" Working Paper
[50] Rector, Robert and Vijay Menon (2018): "Understanding the Hidden $1.1 Trillion
Welfare System and How to Reform It." Backgrounder No. 3284, Heritage Foundation.
[51] Van Parijs, Philippe and Yannick Vanderborght (2017): Basic Income: A Radical Pro-
posal for a Free Society and a Sane Economy. Harvard University Press, Cambridge,
MA.
[52] Voena, Alessandra (2015): "Yours, Mine, and Ours: Do Divorce Laws A¤ect the In-
tertemporal Behavior of Married Couples?" American Economic Review, 105, 2295-
2332.
[53] Wu, Chunzan, and Dirk Krueger (2021): "Consumption Insurance against Wage Risk:
Family Labor Supply and Optimal Progressive Income Taxation," American Economic
Journal: Macroeconomics, 13, 79-113.
47
Appendix to "Rethinking the Welfare State"
1 Transfers to Households in the United States

In this section of the Appendix, we provide a brief description of various means-tested
programs.
1.1 The Earned Income Tax Credit

The EITC is a fully-refundable tax credit that subsidizes low-income working families. The
EITC is a fraction of a family’s earnings until earnings reach a certain threshold. Then, it
stays at a maximum level, and when the earnings reach a second threshold, the credit starts
to decline so that the household does not receive any credit beyond a certain earnings level.
The maximum credits, income thresholds, and the rate at which the credits decline depend
on the household’s tax …ling status (married vs. single) and the number of children. By
design, the EITC only bene…ts working families, and families with children receive a much
larger credit than workers without qualifying children. For 2019, households with earnings
up to 41,000 to 56,000 quali…ed for the EITC. For a married couple or a single parent with
two children, the maximum credit, which applied for earnings around 15 to 20,000$, was
5828$ (a subsidy of more than 25%). The maximum credit for households without children
was much smaller, only 529$. In 2019 about 25 million taxpayers received an average EITC
of $2,476.1
1.2 Child-Related Transfers

Child Tax Credit The CTC provides households a tax credit for each child, indepen-
dent of parents’childcare expenditures and labor market status. Until the 2017 Tax Cuts
and Jobs Act (TCJA), the CTC started at $1,000 per quali…ed children under age 17 and
stayed at this level up to a household income of $75,000 for singles and $110,000 for mar-
ried couples. Beyond this limit, the credit declines by 5% for each dollar earned until it is
completely phased out when the household income is $115,000 for singles and $150,000 for
married couples. The 2017 tax reform increased the maximum credit to 2000$ per child and
allowed households with much higher income to qualify for the maximum credit ($200,000
for single parents and $400,000 for married couples). The CTC is partly refundable: if the
credit exceeds taxes owed, taxpayers can receive up to $1,400 per child, known as the addi-
tional child tax credit (ACTC). To qualify for the ACTC, a household must have minimum
earnings of 2500$.2
The Child and Dependent Care Tax Credit The Child and Dependent Care Tax
Credit (CDCTC) is a non-refundable tax credit that allows parents to deduct a fraction
of their childcare expenses from their tax liabilities. To qualify for the tax credit, both
1
See https://www.eitc.irs.gov/eitc-central/about-eitc/about-eitc for further details.
2
See https://www.irs.gov/newsroom/the-child-tax-credit-bene…ts-eligible-parents and
https://www.taxpolicycenter.org/brie…ng-book/what-child-tax-credit for further details.
1
parents must work. The maximum quali…ed childcare expenditure is $3,000 per child, with
an overall maximum of $6,000. Parents receive a fraction of qualifying expenses as a tax
credit. This fraction starts at 35%, remains at this level up to a household income of $15,000,
and then declines with household income. The lowest rate, which applies to families with a
total household income above $43,000, is 20%.3 Since the CDCTC is not refundable, only
households with positive tax liabilities bene…t from it.
Childcare Subsidies The main program that provides childcare subsidies for low-
income families in the US is the Child Care Development Fund (CCDF). The program was
created as part of the 1996 welfare reform and consolidated an array of programs. To qualify
for a subsidy, parents must be employed, in training, or in school. The program targets low-
income households. In 2010, 1.7 million children (ages 0-13) were served by the CCDF, which
is about 5.5% of all children (ages 0-13) in the US, and the average income of those receiving
a subsidy was about $20,000 (28% of the mean household income) - Guner, Kaygusuz, and
Ventura (2020). Families receiving assistance must make a co-payment, which is about 25%
of childcare costs, while the remaining 75% constituted the subsidy.
1.3 Welfare System

Another group of means-tested programs consist of programs provide cash or in-kind transfers
to poor households, and that are routinely identi…ed with "welfare" system in the US.
Temporary Assistance to Needy Families The TANF was created by the 1996
welfare reform and replaced the Aid to Families with Dependent Children program (AFDC).
Under TANF, the federal government provides a block grant to the states, which use them
to operate their own programs. To receive federal funds, states must also spend some of
their own dollars. The TANF provides monthly cash payments to families, which di¤er
signi…cantly across states. The average maximum monthly payments for a family of three
was 462$ in 2018. The most and least generous states’ payments were 170$ (Mississippi)
and 1039$ (New Hampshire).4 In contrast to the AFDC, the TANF has a 5-year life-time
participation limit and a stronger emphasis on encouraging recipients to work. As a result,
less than 50 percent of TANF spending goes to cash assistance (CBO 2013). The rest pays
for various services for low-income families with children, including child care, transportation
to work, and other types of work-related assistance.
Supplemental Nutrition Assistance Program The SNAP is a federal program that

supports low-income households through electronic bene…t transfer cards that can be used
to buy food. To be eligible, household income, before any of the program’s deductions, must
be at or below 130 percent of the poverty line. For a family of three in 2021, this is $1,810
a month (about $28,200 a year). A family of three with no income receives the maximum
bene…t of 535$ a month. Maximum bene…ts are reduced by 30% for each dollar of monthly
3
See http://www.taxpolicycenter.org/brie…ng-book and https://www.irs.gov/uac/Ten-Things-to-Know-
About-the-Child-and-Dependent-Care-Credit.
4
The Urban Institute’s Welfare Rules Database, TANF Policy Tables, Table II.A.4
https://wrd.urban.org/wrd/tables.cfm. See also Congressional Research Service (2020).
2
household income. On average, SNAP households received about $246 a month in the …scal
year 2020 (Center on Budget and Policy Priorities, 2020).
Supplemental Nutrition Program for Women, Infants, and Children (WIC)

Pregnant, postpartum, and breast-feeding women, infants, and children up to age 5 are
eligible to the WIC if they are poor and an appropriate professional determines them to
be at nutritional risk. An applicant who already receives SNAP, Medicaid, or TNAF is
automatically considered income-eligible for WIC. Applicants who receive no other relevant
means-tested bene…ts must have a gross household income at or below 185 percent of the
federal poverty level (currently $37,296 annually for a family of three) to qualify. WIC
provided an average value of $61.24 in food per participant per month in the …scal year 2016
(Center on Budget and Policy Priorities, 2017).
Supplemental Security Insurance The SSI is a federal program that provides monthly
cash assistance to disabled, blind, or elderly who have little or no income and few assets.
The monthly maximum Federal amounts for 2021 are $794 for an eligible individual, $1,191
for a quali…ed individual with an eligible spouse. In 2019, 79% of payments were for disabled
individuals under age 65 (Social Security Administration, 2020).
Housing Subsidies Several federal programs provide rental assistance to families with
low income. These programs are administered by the Department of Housing and Urban
Development and the U.S. Department of Agriculture and take two forms: i) Public Housing
and ii) Vouchers. In 2017, about 4.6 million households (3.8% of all households in the U.S.)
received some form of federal rental assistance (Mazzara 2017). The amount of aid can
be substantial. Guner, Rauh, and Ventura (2023) calculate that households at the bottom
decile of the income distribution receive about $7,000 per year (about $5,000 for the second
and third lowest deciles).
2 Age-Pro…les: Data and Sample Restrictions

In this section of the Appendix, we presents details on data sources and sample restrictions
for the constructions of age-pro…les presented in Section 2 of the paper. We use the March
Supplement of the CPS from 1980 to 2019 to document how average hourly wages, inequality
of hourly wages and earnings, and labor market statistics (hours and participation) change
over the life cycle. Our measure of inequality is the variance of logs. The analysis is restricted
to household heads and their spouses who are between ages 25 to 60. If a head or a spouse
reports positive earnings or hours, we require that they work at least 520 hours in a year.
To account for top-coded observations, we …t a Pareto distribution to the right tail, as in
Heathcote, Perri and Violante (2010). Finally, we drop observations where the hourly wage
rate (calculated as yearly earnings divided by yearly hours) is less than half of the federal
minimum wage. Given the sensitivity of variance of logs to observations at the lower tail,
we also trim the observations associated with the bottom 0.5% of hourly-wages. These
restrictions are standard in the literature – see Heathcote, Perri, and Violante (2010) and
Huggett, Ventura and Yaron (2011). We calculate total earnings, hours, and hourly wage
3
rates for each individual in the sample. For households, we sum the head and spouse’s
earnings and assign the age of the head to the households.
We then repeat an equivalent procedure using data from the CEX for consumption.
We construct for each household a measure of expenditure of nondurables and services,
which includes food, alcoholic beverages, tobacco products, apparel and services, personal
care, gasoline for transportation, public transportation, household operations, medical care,
entertainment, reading, and education. The de…nition of nondurable consumption follows
Heathcote, Perri, and Violante (2010). The analysis is again based on repeated cross-sections
from the CEX between 1984 and 2019.
Let mj;t;c be any statistic of interest for an age-j individual (or household) at time t; of
cohort c. For example, mj;t;c could be the variance of log hourly wages among j = 30 year
olds in 2000, who are born in c = t j = 1970; i.e. the variance within a (j; t; c)-cell. Since
age, time and cohort are linearly dependent, we construct age pro…les using two approaches.
We …rst consider a time-e¤ects speci…cation by regressing mj;t;c on a set of age and time
(year) dummy variables, i.e.,
0 0
mj;t;c = j Dj + t Dt + "j;t;c ; (1)
where Dj and Dt are a set of age and time dummies. The underlying assumption in the
time-e¤ects speci…cation is that changes in mj;t;c over time are due to time-varying factors
that a¤ect every age (cohort), and once we control for time e¤ects we recover the age pro…les.
Equation (1) is estimated separately for each gender (men and women), marital status (mar-
ried and single), and skill group. For skills, we divide individuals in two groups; skilled (s),
those with at least four years of college education or more, and unskilled (u), with strictly
less than college education. The age pro…les are given by the estimated j values. Then, we
also estimate a cohort-e¤ects speci…cation, given by
0 0
mj;t;c = j Dj + c Dc + j;t;c ; (2)
where Dc is a set cohort dummies. In contrast to equation (1), the underlying assumption in
the cohort-e¤ects speci…cation is that changes in mj;t;c over time re‡ect di¤erences between
younger and older cohorts.
All life-cycle pro…les we use in the benchmark calibration, targeted or non-targeted, are
constructed by controlling for year e¤ects (YE). Here, we show how the life-cycle pro…les
look like when we control for cohort e¤ects (CE) and discuss the properties of the alternative
benchmark economy that uses these pro…les as targeted moments.
Figure A1 shows the variance of log wages by age, skill level, and marital status for
year-e¤ect (dash lines) and cohort-e¤ect (solid lines) speci…cations. The main message that
emerges is that these two speci…cations are broadly consistent. For males, inequality in-
creases more or less linearly along the life cycle. For females, that is not the case; the
increase in inequality slows down after an initial rise around age 35. Quantitatively, there
is a higher increase in inequality over the life cycle under the cohort-e¤ect speci…cation,
which is consistent with estimates provided by Heathcote, Storesletten, and Violante (2010)
and Huggett, Ventura, and Yaron (2011). Figures A2 and A3 show the other life-cycle pro-
…les we report in Section 2 of the paper: the gender wage gap, married female labor force
participation, the variance of log hours worked for females and the correlation of earnings
for husbands and wives. The impact of two alternative speci…cations on these outcomes is
negligible.
4
Married Males Single Males Married Females Single Females
0.60
0.60
0.45
0.45
0.40
0.40
0.50
0.50

0.35
0.35
0.40
0.40
0.30
0.30
0.30
0.30
0.25
0.25
0.20
0.20
0.20
0.20
25 30 35 40 45 50 55 25 30 35 40 45 50 55 25 30 35 40 45 50 55 25 30 35 40 45 50 55
Age Age Age Age
Unskilled, CE Skilled, CE Unskilled, CE Skilled, CE Unskilled, CE Skilled, CE Unskilled, CE Skilled, CE

Unskilled, YE Skilled, YE Unskilled, YE Skilled, YE Unskilled, YE Skilled, YE Unskilled, YE Skilled, YE
Figure A1 - Variance of Log Wages, Males (left) and Females (right), YE and CE
Gender Gap Married Female LFP

1.00
0.85
0.90
0.80
Gender Gap
0.75
0.80
0.70
0.70
0.65
0.60
25 30 35 40 45 50 55 25 30 35 40 45 50 55
Age Age
Unskilled, CE Skilled, CE Unskilled, CE Skilled, CE

Unskilled, YE Skilled, YE Unskilled, YE Skilled, YE
Figure A2 - The Gender Wage Gap (left), LFP of Married Females (right), YE and CE
Var of Log Hours Correlation of Earnings

0.30
0.20
0.25
0.15
0.20
Variance of Log Hours
Correlation
0.15
0.10
0.10
0.05
0.05
0.00
0.00
25 30 35 40 45 50 55 25 30 35 40 45 50 55
Age Age
CE YE CE YE
Figure A3 - Var. of Log Hours, Mar. Females (left), Cor. of Spousal Earnings (right), YE and CE
5
3 Equilibrium
In this section of the Appendix, we de…ne a stationary equilibrium for the model economy.
For all j; let Mj (x; z) = M (x; z) denote the fraction of marriages between age-j; type-x
females and age-j; type-z males, and let ! j (z) = !(z) and j (x) = (x) be the fraction of
single type-z males and the fraction of single type-x females, respectively. The fraction of
type-z males and type-x females are then given by
X
(z) = M (x; z) + !(z); (3)
x2X
and X
(x) = M (x; z) + (x): (4)
z2Z
M
Let let S (x; z; ; ; q; b) be the vector of states that do not change along the life-cycle
for married households, with = ( M M
f;x ; m;z ): For married couples, also summarize the pair
of persistent shocks by ( M M
f;x ; m;z ): Similarly, let Sf
S
(x; Sf;x ; b) and Sm
S
(z; Sm;z )
be the vector of exogenous variables for single females and single males, respectively. In
equilibrium, factor markets clear. The aggregate state of the economy consists of distribution
of households over their types, labor productivity shocks, assets, labor market experience,
and human capital levels. Let the function M M
j (a; h; e; ;S ) denote the number of married
individuals of age j with assets a; human capital level h; female labor market experience
e; current persistent shocks ; and exogenous state S M . The function Sf;j (a; h; e; Sf;x ; SfS ),
for single females, and the function Sm;j (a; Sm;z ; Sm S
), for single males, are de…ned similarly.
Note that household assets, a; and female human capital levels, h; are continuous decisions.
Let a 2 A = [0; a] and H = [0; h] be the sets of possible assets and female human capital
levels. Let the set for possible values of the market experience be denoted by E = [0; e]: By
construction, M (x; z); the number of married households of type (x; z); must satisfy for all
j
X Z
M M
M (x; z) = j (a; h; e; ;S )dh da de d :
; ;q;b A H
Similarly, the fraction of single females and males must be consistent with the corresponding
measures Sf;j and Sm;j , i.e. for all ages, we have
XZ
S S
(x) = f;j (a; h; e; ; Sf )dh da de d ;
;b A H E
and XZ
S S
!(z) = m;j (a; ; Sm )da d :
A
For married couples, let Mb (x; z) be the fraction of type-(x; z) couples who have child-
bearing type b (where b 2 f0;
P 1; 2g denotes no children, earlyS childbearing and late child-
bearing, respectively), with b M b (x; z) = 1. Similarly, let b (x) be the fraction of type-x
6
P
single females who have childbearing type b; with b Sb (x) = 1: Let the decision rules as-
sociated with the dynamic programming problems outlined in Section 3.5 of the paper be
denoted by aSm (a; Sm;z ; Sm S S
; j) and lm S
(a; Sm;z ; Sm ; j) for single males, by aSf (a; h; e; Sf;x ; SfS ; j)
and lf (a; h; e; f;x ; Sf ; j) for single females, and by aM (a; h; e; ; S M ; j), lfM (a; h; e; ; S M ; j)
S S S
M
and lm (a; h; e; ; S M ; j) for married couples. Finally, let the functions hS (a; h; e; ; SfS ; j)
and hM (a; h; e; ;S M ; j) describe the next period’s human capital for a single and married
female, respectively. They are de…ned as
hM (a; h; e; ;S M ; j) = H(x; h; lfM (a; h; e; ; S M ; j); e);
and
hS (a; h; e; ; SfS ; j) = H(x; h; lfS (a; h; e; ; SfS ; j); e);

where H is the human capital accumulation function. Let f:g denote the indicator function.
Summarize the transition functions for persistent shocks by M ( 0 j ); Sf ( 0 j ) and Sm ( 0 j )
and the initial draws for permanent shocks by M ( ); Sf ( ); and Sm ( ):
In equilibrium, the distribution functions M M
j (a; h; e; ;S ) ,
S S S
f;j (a; h; e; f;x ; Sf ), and
S S S
m;j (a; m;z ; Sm ) must obey the following recursions:
Married agents
Z
M 0 0 0 0 M M M
j (a ; h ; e ; ;S ) = ( 0j ) j 1 (a; h; e; ;S M ) (5)
faM (a; h; e; ; S M ; j 1) = a0 ; hM (a; h; e; S M ; ; j 1) = h0 gdh da de d ;
for j > 1 with

e; if lfM (a; h; e; ; S M ; j 1) = 0
e0 = :
e + 1; otherwise
For j = 1;
M M (x; z) M
b (x; z)
M
( )( (qjz) if a = 0; e = 0; = 0; h = $m (x; 1);
1 (a; e; h; ;S M ) = ;
0; otherwise
where $m (x; 1) is a function that maps female types their initial human capital, (qjz) is
the fraction of households that draw q (given z); and is the probability of drawing :
Single female agents
Z
S 0 0 0 0 S
f;j (a ; h ; e ; ; SfS ) = S 0
f( j ) f;j 1 (a; h; e; ; SfS ) (6)
faSf (a; h; e; ; SfS ; j 1) = a0 ; hS (a; h; e; ; SfS ; j 1) = h0 gdh da de d ;
for j > 1; with again
e; if lfS (a; h; e; Sf;x ; SfS ; j 1) = 0

e0 = ;
e + 1; otherwise
7
and
S (x) Sf ( ) Sb (x) if e = 0; = 0; h = $f (x; 1)

f;1 (a; h; e; ; SfS ) = :
0; otherwise
Single male agents
Z
S S S 0 S S
m;j (a; ; Sm ) = f( j ) m;j 1 (a; ; Sm ) faSm (a; ; Sm
S
;j 1) = a0 gda d ; (7)
for j > 1; and
S S $m (z; 1) Sm ( ) if a = 0; =0
m;1 (a; ; Sm )= :
0; otherwise
Given distribution functions M M
j (a; h; e; ;S ) ,
S
f;j (a; h; e;
S S
f;x ; Sf ), and S
m;j (a;
S S
m;z ; Sm ),
the aggregate capital stock is given by
X XZ XZ
M M S S
K = j[ a j (a; h; e; ; S )dh da de d + a m;j (a; ; Sm )da d (8)
j SM S
Sm
XZ
S
+ a f;j (a; h; e; ; SfS )dh da de d ]:
SfS
The skilled labor input, Ls , is given by
X X Z
M
Ls = j[ (h exp( + )lfM (a; h; e; ; S M ; j) j (a; h; e; ; S M )dh da de d
j S M :x=s
X Z
M M
+ ($(z; j) exp( + )lm (a; h; e; ; S M ; j) j (a; h; e; ; S M )dh da de d
S M :z=s
X Z
S S S S
+ $(z; j) exp( + )lm (a; ; Sm ; j) m;j (a; ; Sm )da d (9)
S :z=s
Sm
X Z
S
+ h exp( + )lfS (a; h; e; ; SfS ; j) f;j (a; h; e; ; SfS )dh da de d ]:
SfS :x=s
In turn, the (total) unskilled labor input, is given by
8
X X Z
M
Lu = j[ (h exp( + )lfM (a; h; e; ; S M ; j) j (a; h; e; ; S M )dh da de d
j S M :x=u
X Z
M M
+ ($(z; j) exp( + )lm (a; h; e; ; S M ; j) j (a; h; e; ; S M )dh da de d
S M :z=u
X Z
S S S S
+ $(z; j) exp( + )lm (a; ; Sm ; j) m;j (a; ; Sm )da d (10)
S :z=u
Sm
X Z
S
+ h exp( + )lfS (a; h; e; ; SfS ; j) f;j (a; h; e; ; SfS )dh da de d ];
SfS :x=u
Furthermore, unskilled labor used in the production of goods, Lu;g , equals the total supply
of unskilled labor net of its usage in the production of childcare services:
X Z
M
Lu;g = Lu [ j flfM gD(x; z; b; j) j (a; h; e; ; S M ; j)dh da de d
S M
X Z
S
+ j flfS gD(x; b; j) f;j (a; h; e; ; SfS ; j)dh da de d ]:
SfS
In equilibrium, total taxes must cover government expenditures, G, total government

spending and total transfers, T R, i.e.,
X XZ
M
G + TR = j[ T M (I; K(:)) j (a; h; e; ; S M )dh da de d (11)
j SM
XZ
S
+ T S (I; 0) m;j (a;
S
; Sm ) da d
S
Sm
XZ
S
+ T S (I; K(:)) f;j (a; e; h; ; SfS ) dh da de d ] + k rK;
SfS
where I represents a household’s total income and K the number of children as as de…ned in
the description of the individual and household problems in Section 3.5 of the paper. The
aggregate transfers are given by
X XZ
TR = j[ T RM (I; K(:); D) M M
j (a; h; e; ; S )dh da de d
j SM
XZ
S S S
+ T Rm (I) m;j (a; ; Sm ) da d
S
Sm
XZ
S
+ T RfS (I; K(:); D) f;j (a; e; h; ; SfS ) dh da de d ];
SfS
9
where D stands for childcare expenditures, as de…ned in Section 3.5 of the paper.
Finally, the social security budget must balance
X XZ XZ
M
j[
M M
p (x; z) j (a; h; e; 0; S )dh da de + pSf (x) Sf;j (a; e; h; 0; SfS )dh da de
j JR SM SfS
XZ
S
+ pSm (z) S
m;j (a; 0; Sm ) da] (12)
S
Sm
= p [ws Ls + wu Lu ]:
Equilibrium De…nition For a given government consumption level G, social security

bene…ts pM (x; z), pSf (x) and pSm (z); tax functions T S (:), T M (:), a payroll tax rate p , a cap-
ital tax rate k , transfer function T RfS (:); T Rm S
(:); T RM (:); and an exogenous demographic
structure represented by (z); (x); M (x; z), and j ; a stationary equilibrium consists of
prices r and (ws ; wu ); aggregate capital (K), aggregate labor (Ls ; Lu ; Lu;g ), household deci-
sion rules aSm (a; Sm;z ; Sm S S
; j) and lm S
(a; Sm;z ; Sm ; j) for single males, by aSf (a; e; h; Sf;x ; SfS ; j)
and lfS (a; e; h; Sf;x ; SfS ; j) for single females, and by aM (a; h; e; ; S M ; j), lfM (a; h; e; ; S M ; j)
and lm M
(a; h; e; ; S M ; j) for married couples., and distribution functions M j (a; e; h; ;S ) ,
M
S S S S S S
f;j (a; e; h; f;x ; Sf ), and m;j (a; m;z ; Sm ); such that
1. Given tax and transfer rules, and factor prices, the decisions of households are optimal.
@F (K;Lg ) @F (K;Lg )
2. Factor prices are competitively determined; i.e. ws = @Ls
, wu = @Lu;g
and
@F (K;Lg )
r= @K k.
3. Factor markets clear; i.e. equations (8) and (9) hold.
4. The functions M S S
j , f;j , and m;j are consistent with individual decisions, i.e. they are
de…ned by equations (5), (6), and (7).
5. The government and social security budgets are balanced; i.e. equations (11) and (12)
hold.
4 Model Inputs and Calibration

4.1 Demographics
The model period is a year. The population grows at the annual rate of 1.1%, the average
values for the U.S. economy between 1960-2000. We determine the distribution of individuals
by productivity types for each gender, using data from the 2008 American Community Survey
(ACS). We consider all household heads or spouses between ages 30 and 39 and for each
gender calculate the fraction of population in each education cell. For the same age group,
the distribution of married working couples, as shown in Table A1. Given the fractions of
individuals in each education group and the fractions of married households, we calculate the
implied fractions of single households. The resulting values are reported in Table A2. About
10
74% of households consist of married households, while the rest (about 26%) are single.
Since we assume that the distribution of individuals by marital status is independent of age,
we use the 30-39 age group in the calibration. This age group captures the marital status of
recent cohorts during their prime-working years, while being at the same time representative
of older age groups.
Table A1: Distribution of Married Working Households by Type

Females
Males Unskilled Skilled
Unskilled 51.37 12.81
Skilled 8.93 26.90
Note: Entries show the fraction of marriages out of the total married pool, by wife and husband educa-
tional categories. The data used is from the 2008 ACS, ages 30-39. Entries add up to 100.
Table A2: Fraction of Agents by Type, Gender and Marital Status

Males Females
All Married Singles All Married Singles
Unskilled 65.38 48.19 17.19 62.23 44.03 18.21
Skilled 34.62 26.51 8.11 37.77 29.10 8.66
Note: Entries show the fraction of individuals in each educational category, by marital status, constructed
under the assumption of a stationary population structure.
4.2 Children
In the model each single female and each married couple belong to one of three groups:
without children, early child bearer and late child bearer. We use information on the age of
last birth of mothers by skill to determine who is in each category. The unskilled early child
bearers have all children at age 1 (age 25). Skilled early-child bearers have children at age 1
(25) and at age 3 (27). Late child bearers have their children at ages 8 and 10, corresponding
to ages 32 and 34. This structure captures the fact that births occur within a short time
interval; between 25 and 29 for unskilled and between 30 and 34 for skilled households in
the 2008 CPS June (Fertility) Supplement.5 From the 2008 CPS June Supplement, we also
calculate the fraction of 40 to 44 years old single (never married or divorced) females with
zero live births. This provides us with a measure of lifetime childlessness. Then we calculate
the fraction of all single women above age 25 with a total number of two live births who were
below age 30 at their last birth. This fraction gives us those who are early child bearers,
and the remaining fraction are assigned as late child bearers. The resulting distribution is
shown in Table A3.
We follow a similar procedure for married couples, combining data from the CPS June
Supplement and the U.S. Census. For childlessness, we use the larger sample from the U.S.
5
The CPS June Supplement provides data on the total number of live births and the age at last birth for
females, which are not available in the U.S. Census.
11
Census.6 The Census does not provide data on total number of live births but the total
number of children in the household is available. Therefore, as a measure of childlessness
we use the fraction of married couples between ages 35-39 who have no children at home.7
Then, using the CPS June supplement we look at all couples above age 25 in which the
female had a total of two live births and was below age 30 at her last birth. This gives us the
fraction of couples who are early child bearers, with the remaining married couples labeled
as the late ones. Table A4 shows the resulting distributions. Table A5 displays the number
of children for single mothers by skill, and the corresponding ones for married couples.
Table A3: Childbearing Status, Single Females

Childless Early Late
Unskilled 29.27 57.42 13.31
Skilled 54.63 28.17 17.20
Note: Entries show the distribution of childbearing among single females, using data from the
CPS-June supplement.
Table A4: Childbearing Status, Married Couples

Childless Early
Females Females
Male Unskilled Skilled male Unskilled Skilled
Unskilled 9.22 13.17 Unskilled 63.46 40.58
Skilled 9.89 11.51 Skilled 45.88 26.95
Note: Entries show the distribution of childbearing among married couples. For childlessness,
data used is from the U.S. Census. For early childbearing, the data used is from the CPS-June
supplement. Values for late childbearing can be obtained residually for each cell.
Table A5: Fertility Di¤erences

Singles Married
Females
Male Unskilled Skilled
Unskilled 2.21 Unskilled 2.34 2.05
Skilled 1.82 Skilled 2.33 1.98
Note: Entries show, conditional on having children, the total number of children di¤erent
types of households have by age 40-44. The authors’ calculations from the 2008 CPS-June
supplement.
6
The CPS June Supplement is not particularly useful for the calculation of childlessness in married
couples. The sample size is too small for some married household types for the calculation of the fraction of
married females, aged 40-44, with no live births.
7
Since we use children at home as a proxy for childlessness, we use age 35-39 rather than 40-44. Using
ages 40-44 generates more childlessness among less educated people. This is counterfactual, and simply
results from the fact that less educated people are more likely to have kids younger, and hence these kids
are less likely to be at home when their parents are between ages 40-44.
12
Childcare Costs We use the U.S. Bureau of Census data from the Survey of Income
and Program Participation (SIPP) to calibrate childcare costs. We estimate a relation that
represents the relation between the average age of children at home and per-child childcare
costs, conditional on mother’s skills and marital status. We estimate:
b t; mar) = amar + bmar ln(t);
d(x; x x
where mar 2 fM; Sg stands for marital status, and t is the average age of children at home.
b t; mar); re‡ects e¤ective spending, so
The childcare spending per children in the data, d(x;
captures di¤erences among household in access to informal care or quality of childcare chosen.
Figure A4 (right panel) shows the estimated values. Our estimates imply that childcare costs
are larger for skilled mothers and decline fast as children age. The annual rate of decline is
about 11-12% (10-11%) when the child age is …ve for skilled (unskilled) mothers.
The childcare costs of a married couple where the wife is of skill x are given by wu dM (x; t) =
b t; M ) for each t, while for a single woman are given by wu dS (x; t) = d(x;
d(x; b t; S). The result-
ing values for e¢ ciency units are scaled so that the total childcare expenditure for children
between ages 0 to 5 is in line with the data. As documented in Guner, Kaygusuz, and
Ventura (2022), the total yearly cost for employed mothers, who have children between 0
and 5 and who make childcare payments, was about $6,414.5 in 2005, which is about 10%
of average household income. In the benchmark economy, this choice of parameter values
results in 1.1% of the total labor input being used to produce childcare services. This is in
line with the share of employment in the childcare sector in the U.S., which was about 1.1%
in 2012.8
0.90
1.50
0.80
1.00
Childcare Costs
0.70
0.50
0.60
0.50
0.00
0 5 10 15 0 5 10 15
Age of Youngest Child Age
Unskilled Skilled Married, skilled Married, unskilled Single, skilled Single, unskilled
Figure A4 - LFP of Mar. Fem., by age of the youngest child (left); Childcare Costs per Child (right)
4.3 Taxes
4.3.1 Income Taxes
The tax function parameters are taken from Guner, Kaygusuz and Ventura (2022), who
follow Guner, Kaygusuz and Ventura (2014) and estimate e¤ective tax rates as a function
8
Total employment in childcare services (NAICS 6244) was about 1.6 million in 2012. This num-
ber is the sum of total paid employment and the number of establishments without paid employees. See
http://thedataweb.rm.census.gov/TheDataWeb_HotReport2/econsnapshot/2012/snapshot.hrml?NAICS=6244.
13
of reported income, marital status, and the number of children. The data is tax-return,
micro-data from Internal Revenue Service for the year 2000 (Statistics of Income Public Use
Tax File).
Since the EITC, CTC and CDCTC are explicitly modeled in the benchmark economy,
tax liabilities in the absence of these credits are considered. To this end, let I stand for
multiples of mean household income in the data and denote by e t(I) the corresponding tax
liabilities after any tax credits. Tax credits reduce the tax liability …rst to zero and if there
is any refundable credit left, the household receives a transfer. Let credit(I) be the total
credits without any refunds, which we can identify in the IRS micro tax data. Taxes in the
absence of credits is then given by t(I) = e t(I) + credit(I): The incomes tax functions, i.e.
S M
T (I; k) and T (I; k);take the following form
(I) = 1 I ;
where I is measured in multiples of mean household income, (I) is the average tax rate,
parameter determines the progressivity of taxes and determines the taxes at the mean
household income (I = 1): Parameters and depend on marital status and the number of
children. The total tax liabilities amount to (I) I mean household income.
Estimates for and are contained in Table A6. Further details are provided in Guner,
Kaygusuz and Ventura (2022). Guner, Kayusuz and Ventura (2014) show that this functional
form does a great job matching average and marginal tax rates in the data. We estimate tax
functions for households with zero and two children (and assign the number of children from
Table A5 by rounding the numbers to the nearest integer). Figure A5 (left panel) displays
estimated average and marginal tax rates for di¤erent multiples of household income.
Table A6: Tax Functions
Estimates Married Single
(no child) (2 child.) (no child) (2 child.)
0.9024 0.9078 0.8815 0.9227
0.0569 0.0596 0.0356 0.0351
Note: Parameter estimates from Guner, Kaygusuz, and Ventura (2022).
0.15
0.20
0.15
0.10
Average Tax Rate
0.10
0.05
0.05
0.00
0.00
0 2 4 6 8 10 0 .5 1 1.5
Household Income (normalized)
Married, 2 children Single Female, 2 Children Single Male, no child
Married, 2 children Single, two children
Figure A5 - Average Taxes (left); Welfare Payments (right)
14
4.3.2 Social Security and Capital Taxation
We calculate p = 0:086; as the average value of the social security contributions as a fraction
of aggregate labor income for 1990-2000 period.9 Using the 2008 ACS, we calculate total
Social Security bene…ts for all single and married households.10 Tables A7 shows Social
Security bene…ts, normalized by the level corresponding to single males of the lowest type,
pSm (z1 ). We treat pSm (z1 ) as a free parameter, and determine all other bene…t levels according
to Table A7. Then, given p , choose pSm (z1 ) to balance the budget for the social security
system. Hence, while the relative values social security bene…ts come from the data, the
absolute level of one, pSm (z1 ); is adjusted to balance the budget of the system. The implied
value of pSm (x1 ) for the benchmark economy is about 18.1% of the average household income
in the economy.We use k to proxy the U.S. corporate income tax. We estimate this tax
rate as the one that reproduces the observed level of tax collections out of corporate income
taxes after the major reforms of 1986. Such tax collections averaged about 1.92% of GDP
for 1987-2000 period. Using the technology parameters we calibrate in conjunction with our
notion of output (business GDP), we obtain k = 0:097.
Table A7: Social Security Bene…ts

Single Married
Females
Unskilled Skilled Males Unskilled Skilled
Males 1 1.166 Unskilled 1.764 1.911
Females 0.888 0.995 Skilled 1.981 2.093
Note: Entries show Social Security bene…ts, normalized by the mean Social Security income of
the lowest type male, using data from the 2008 ACS.
4.4 Welfare State

Transfers, T RfS (I; k; D); T Rm
S
(I); and T RM (I; k; D); consist of three components. The …rst
component is the Earned Income Tax Credit (EITC). The second part is child-related trans-
fers, which consists of Child Tax Credit (CTC), the Child and Dependent Care Tax Credit
(CDCTC), and childcare subsidies. The …nal component is the means-tested transfers.
4.4.1 Earned Income Tax Credits (EITC)

We model all tax credits as they appear in 2004 tax code. Since we represent all variables
as a fraction of the mean household income, in the absence of any changes in the tax code,
the reference year is not critical. The Earned Income Tax Credit is a fully refundable tax
credit that subsidizes low income working families. The EITC amounts to a …xed fraction
9
The contributions considered are those from the Old Age, Survivors and DI programs. The Data comes
from the Social Security Bulletin, Annual Statistical Supplement, 2005, Tables 4.A.3.
10
Social Security income is all pre-tax income from Social Security pensions, survivors bene…ts, or perma-
nent disability insurance. Since Social Security payments are reduced for those with earnings, we restrict
our sample to those above age 70. For married couples we sum the social security payments of husbands and
wives.
15
of a family’s earnings until earnings reach a certain threshold. Then, it stays at a maximum
level, and when the earnings reach a second threshold, the credit starts to decline, so that
beyond a certain earnings level the household does not receive any credit. The amount of
maximum credits, income thresholds, as well as the rate at which the credits declines depend
on the tax …ling status of the household (married vs. single) as well as on the number of
children. In 2004, for a married couple with 0 (2 or 3) children, the EITC started at $2 ($10)
and increased by 7.6 (39.9) cents for each extra $ in earnings up to a maximum credit of
$3,900 ($4,300). Then the credit stays at this level until the household earnings are $7,375
($15,025). After this level of earnings, the credit starts declining at a rate of 7.6 (21) cents
for each extra $ in earnings until it becomes zero for earrings above $12,490 ($35,458). The
formulas for a single household with 0 (2 or 3) children are very similar. We calculate the
level of EIT C as a function of earnings with the following formula,
EIT C = maxfCAP - maxfslope1 (bend1 earnings); 0g

maxfslope2 (earnings bend2 ); 0g; 0g;
where CAP; the maximum credit level, bend1 and bend2 ; the threshold levels, and slope1
and slope2 , the rate at which credit increase and decline are given by ( as a fraction of mean
household income in 2014):
CAP slope1 bend1 slope2 bend2

Married
No ch. 0.006 0.076 0.085 0.076 0.122
2 or 3 ch. 0.071 0.399 0.178 0.21 0.248
Single
No ch. 0.006 0.076 0.085 0.076 0.105
2 or 3 ch. 0.071 0.399 0.178 0.21 0.232
Figure A6 (left panel) shows the EITC as a function of household income and the tax …ling
status.
EITC Child Tax Credit and Childcare Tax Credit

0.08
0.04
0.06
0.03
Credits, % of mean income
EITC, % of mean income
0.04
0.02
0.02
0.01
0.00
0.00
0 .2 .4 .6 .8 1 0 1 2 3
Income (normalized) Income (normalized)
Married, 2 children Single, two children CTC CDCTC
Figure A6 - The Earned Income Tax Credit (left); Potential CTC and CDCTC (right)
16
4.4.2 Child Tax Credits
Child credits operate as a means-tested transfer to households with children. If a household’s
income is below a certain limit, IbCT C ; then the potential credit is dCT C = $1; 000 per child
in 2004. If the household income is above the income limit, then the credit amount declines
by 5% for each additional dollar of income. In the current tax code, IbCT C is $110,000 for
a married couple and $75,000 for singles. As a result, a married couple with two children
whose total household income is below $110,000 has a potential child credit of $2,000, a
household with two children whose total household income is $120,000 can only get $1,500.
The child credit becomes zero for married couples (singles) whose total household income
is above $150,000 ($115,000). As the CTC is not fully refundable, the actual CTC that a
household gets depends on the total tax liabilities of the household and other child-related
credits that the household might qualify.
For a household with income level I (again indicated as a multiple of mean household
income in the economy) and k children; the potential CT C is given by
CT Cpotential (I) = maxf[k 0:0165 max(I IbCT C , 0) 0:05]; 0g; (13)
with
1:819; if married …ling jointly
IbCT C = ;
1:240; if single
where again the maximum amount of credit per child, 0.0165, and income limits, 1.819 and
1.240, are in multiples of mean household income in the U.S. in 2004. Both the CTC and
the CDCTC are non-refundable, as a result, how much of the potential credit a household
actually gets depends on its total tax liabilities and total tax credits (CTC plus CDCTC). Let
Creditpotential (I) = CT Cpotential (I) + CDCT Cpotential (I) and T axes(I) be the total potential
tax credits and the tax liabilities of the household. Then,
8
>
> CDCT Cpotential (I); if T axes(I) > Creditpotential (I)
>
>
< maxfT axes(I) CDCT Cpotential (I); 0g; if T axes(I) < Creditpotential (I)
CDCT Cactual (I) = and CDCT Cpotential (I) > T axes(I) ;
>
>
>
> CDCT Cpotential (I); if T axes(I) < Creditspotential (I)
:
but CDCT Cpotential (I) < T axes(I)
and
8
>
> CT Cpotential (I); if T axes(I) > Creditspotential (I)
>
>
< 0; if T axes(I) < Creditspotential (I)
CT Cactual (I) = and CDCT Cpotential (I) > T axes(I)
>
>
>
> T axes(I) CDCT Cpotential (I); if T axes(I) < Creditspotential (I)
:
but CDCT Cpotential (I) < T axes(I)
Hence, if the tax liabilities of a household are larger than the total potential credit
implied by the CTC and the CDCTC, the household receives the full credit and its tax
liabilities are reduced by CT Cpotential + CDCT Cpotential : If the total potential credits are
larger than tax liabilities, then the household only receives a credit up to its tax liabilities.
As a result, the households with low tax liabilities do not bene…t from the CTC or CDCTC.
17
This is partially compensated by the Additional Child Tax Credit (ACTC), which gives a
household additional tax credits if its potential child tax credit is higher than the actual
child tax credits it receives. In order to qualify for the ACTC, however, a household must
have earnings above $10,750. Thus, a household with very low earnings does not qualify for
the ACTC. Given CT Cactual and CT Ccredit ; the ACTC is calculated as
8
< minfmax[(earnings 0:178); 0] 0:15; CT Cpotential (I) CT Cactual (I)g
ACT C(I) = if CT Cactual (I) CT Ccredit (I) :
:
0; otherwise
4.4.3 Childcare Credits

All households with positive income can qualify for the Child and Dependent Care Tax Credit
(CDCTC), or, as we refer in the paper, for childcare credits. Potential childcare credits are
calculated in two steps, using the total childcare expenditures of the household, a cap, and
rates that depend on household income. First, for each household, a childcare expenditure
that can be claimed against credits is calculated. This expenditure is simply the minimum
of the earnings of each parent in the household, a cap and actual childcare expenditures.
The cap is set $3,000 and $6,000 for households with one child and with more than one child
in 2004. Second, each household can claim a certain fraction of this quali…ed expenditure as
a tax credit. This fraction starts at 35%, and declines by household income by 1% for each
$2,000 above $15,000 until it reaches 20%, and then remains constant at this level. For a
married couple with k children; the quali…ed expenditure is calculated as follows
Expense = minfdCDCT C minfk; 2g; earnings1 ; earnings2 ; dg;
where earnings1 and earnings2 are the earnings of the household head and his/her spouse
and d is the child care expenditure (net of any childcare subsidy that a household might
qualify). Note that a married couple household can have quali…ed expenses only if both
the husband and the wife have non-zero earnings. The child care expenditures for the
calculation of the childcare credits are capped at dCDCT C per child per year, with a maximum
of 2 dDCCT C .
For a single female household, the equivalent formula is given by
Expense = minfdCDCT C minfk; 2g; earnings; dg:
In 2004, dCDCT C was $3,000, i.e. maximum quali…ed expenditure for households with more
than 1 child was capped at $6,000. In multiples of mean household income in the U.S.
($60,464 in that year), dCDCT C was equal to 0:0496; i.e. about 5% of mean household
income in the US. A household, however, only receives a fraction CDCT C (I) of quali…ed
expenses. The rate, CDCT C ; is a declining function of household income. It is set at 35% for
households whose income is below $15,000 (IbCDCT C ); and after this point the rate declines
by 1% for each extra $2,000 that the household earns down to a minimum of 20%. Hence,
the potential CDCT C that a household can receive is then given by
CDCT Cpotential (I) = Expense CDCT C (I), (14)
18
with
0:35; if I IbCDCT C
CDCT C (I) = IbCDCT C
;
0:35 minf[integer( I 0.033
) + 1] 0.01, 0.15}, otherwise
where IbCDCT C is equal to 0:248 is in multiples of mean household income in the U.S. in 2004.
Figure A6 (right panel) illustrates the sum of CDCT Cpotential (I) and CT Cpotential (I).11
4.4.4 Childcare Subsidies

We assume that the childcare subsidies in the model economy re‡ect the childcare subsidies
provided by the Children Child Care and Development Fund (CCDF) in the US. Following
Guner, Kaygusuz and Ventura (2022), we set = 0:75 and choose Ib such that the poorest
5.5% of families with children receive a subsidy from the government. This procedure sets Ib
at about 24.2% of mean household income in the benchmark economy. In the main policy
experiments that we consider, we make the childcare subsides universal by setting Ib to an
arbitrarily large number.
4.4.5 Means-Tested Transfers

The means tested transfers are taken from Guner, Kaygusuz and Kaygusuz (2022), who use
the 2004 wave of the Survey of Income and Program Participation (SIPP) to approximate a
welfare schedule as a function of labor earnings for di¤erent household types. The "e¤ective
transfer function" (conditional on marital status and the number of children) takes the
following form
! 0 if I = 0
W (I) = ;
maxf0; ! 1 ! 2 Ig if I > 0
where ! 0 is the transfers for a household with zero income and ! 2 is the bene…ts reduction
rate and I is reported in multiples of mean household income. To determine ! 0 ; the average
amount of welfare payments for households with zero non-transfer income is used. Then
an OLS regression of welfare payments on household non-transfer income is estimated to
determine 0 and 1 . In Table A8 shows the estimated values of ! 0 ; 1 and 2 and Figure
A5 (left panel) shows the welfare payments as a function of household income. Further
details are provided in Guner, Kaygusuz and Ventura (2022).
Table A8: Welfare System

Estimates Married Single Female Single Male
(no child) (2 child.) (no child) (2 child.) (no child)
!0 0.063 0.090 0.090 0.116 0.075
!1 0.023 0.043 0.044 0.101 0.032
!2 -0.017 -0.033 -0.042 -0.091 -0.028
Note: Parameter estimates from Guner, Kaygusuz, and Ventura (2022).
11
The simulations for CDCT Cpotential (I) in Figure A4 are done under the assumption th at at each income
level, the husband and the wife earns 60% and 40% of the household income, respectively, and the households
spend 10% of their income on childcare.
19
4.5 Heterogeneity
There are 2 education types of males, corresponding to educational attainment levels less than
college (u), and college or more (s). We use the March Supplement of the CPS from 1980
to 2019 to calculate age-e¢ ciency pro…les for each male type. For the benchmark economy,
we construct age pro…les for di¤erent outcomes from cross-sectional data by removing year
e¤ects, as detailed in Section 2 of the paper. Within a skill group, e¢ ciency levels correspond
to mean weekly wage rates, which we construct using annual wage and salary income and
weeks worked, normalized by the mean weekly wages for all males and females between ages
25 and 64. Figure A7 (left panel) shows the third-degree polynomials that we …t to the wage
data. In the quantitative exercises, the male e¢ ciency units, $m (z; j); correspond to these
…tted values.
There are also 2 education types for females. Table A9 reports the initial (age 25) e¢ -
ciency levels for females together with the initial male e¢ ciency levels and the corresponding
gender wage gap. We use the initial e¢ ciency levels for females to calibrate their initial hu-
man capital levels, h1 = $f (x; 1): After age 25, the human capital level of females evolves
endogenously according to
h0 = H(x; h; l; e) = exp [ln h + e

x (l) x (1 (l))] ; x 2 X = fu; sg;
where e stands for labor market experience and (:) is an indicator function that is 1 if hours
worked are positive and zero otherwise. Parameter ex is experience-skill growth rate and x
stands for the depreciation rate.
We calibrate the values for x and ex as follows. First, we select ex so that if a female
of a particular education type works in every period, her wage pro…le has exactly the same
shape as a male of the same type. This procedure takes the initial gender di¤erences as
given, and assumes that the wage growth rate for a female who works full time will be the
same as for a male worker with the same level of experience; hence, it sets ex values equal
to the growth rates of male wages at each age. Figure A7 (right panel) shows the calibrated
values for ex : We then select two values of x so that we match the level of gender gap for
skilled and unskilled women by age 25-35 as closely as possible.12
Table A9: Initial Productivity Levels, by Type and Gender

$m (1; z) $f (1; x) $f (1; x)=$m (1; z)
Skilled 0.88 0.81 0.92
Unskilled 0.69 0.56 0.80
Note: Entries are the productivity levels of males and females, ages 25, using 1980-2019 data
from the CPS March Supplement. These levels are constructed as weekly wages for each type.
12
We target the gender gap in hourly wages all married females in the model. We impute wages for females
who do not participate using a standard Heckman (1979) selection correction. For the population equation
for wages, we assume a standard Mincer equation, i.e. log wages of women depend on years of education,
age, and age squared. For the selection equation, we assume that the probability of participation in the
labour market for a female depends on her marital status, number of children younger than age 5, and the
variables in the population equation.
20
2.00
0.08
0.06
1.50
Normalized Wages
Wage Growth
0.04
1.00
0.02
0.00
0.50
25 30 35 40 45 50 55 60 65 25 30 35 40 45 50 55 60 65
Age Age
Unskilled Unskilled, fitted Skilled Skilled, fitted Unskilled Skilled
Figure A7 - Age-Labor Productivity Pro…les, Males (left); Female Human Capital Growth (right)
4.6 Preferences and Technology

In this section of the Appendix, we provide further details on how we assign parameter values
to the endowment, preference, and technology parameters of the benchmark economy. There
are three utility-function parameters to be determined: the intertemporal elasticity of labor
supply ( ), the parameter governing the disutility of market work for males and females,
Bm and Bf , and the disutility shock of market work for married females, : We set the
Frisch elasticity parameter to 0.2. This is on the low side of recent available estimates, but
via other choices in our economy, the macro elasticity is broadly consistent with estimates.
Given , we select the parameter Bf and Bm to reproduce average market hours per worker
observed in the data, about 42.7% and 37.0% of available time for males and females in
2008.13 Finally, the disutility shocks are speci…ed as L = exp( ) and H = exp(+ ).
The parameter is set so as to reproduce the observed variance of log-hours of married
females at age 40 (0.127 in the data). As it is the standard in the literature, we select the
discount factor , so that the steady-state capital to output ratio matches the value in the
data (2.93).
Utility costs associated to joint work allow us to capture the residual heterogeneity among
couples, beyond heterogeneity in endowments and childbearing status, that is needed to ac-
count for the observed heterogeneity in participation choices. We assume that the utility cost
parameter of joint participation is distributed according to a gamma distribution, approx-
imated on a discrete grid, with parameters kz and z . Thus, conditional on the husband’s
type z,
1 exp( q= z )
q (qjz) q kz ;
(kz ) kz z
where (:) is the Gamma function. This procedure allows us to exploit the information
contained in the di¤erences in the labor force participation of married females as their own
13
The numbers are for people between ages 25 and 54 and are based on data from the CPS. We …nd mean
yearly hours worked by all males and females by multiplying usual hours worked in a week and number of
weeks worked. We assume that each person has an available time of 5; 000 hours per year.
21
wage rate changes with skill. In this way, we indirectly control the ’slope’of the distribution
of utility costs, which is potentially key in assessing the e¤ects of changing incentives for
labor force participation.
Table A10: Labor Force Participation of Married Females, 25-54

Females
Unskilled 69.1 85.2
Skilled 64.8 73.3
Note: Each entry shows the labor force participation of married females ages 25 to 54, calculated
from the 2008 ACS. The outer row shows the weighted average for a …xed male or female type.
Using the Census data, we calculate that the employment-population ratio of married
females between ages 25 and 54, for each of the educational categories de…ned earlier.14 Table
A11 shows the resulting distribution of the labor force participation of married females by
the productivities of husbands and wives for married households. The aggregate labor force
participation for this group is 71.8%, and it increases from 68.2% for the unskilled group
to 77.4% for the skilled. Our strategy is then to select the two parameters governing the
gamma distribution, for every husband type, so as to reproduce each of the rows in Table
A10 as closely as possible. This process requires estimating four parameters (i.e. a pair
( ; k) for each husband educational category). Given the estimated values for kz and z ; we
determine the loading factors #x (tmin ) so that the model is consistent with the participation
rate of mothers by the age of their youngest child present at home, shown in Figure A4 (left
panel). To compute the participation rate of married females by skill by the age of their
youngest child at home, we use data from the 2008 ACS.
Finally, we set the capital share to = 0:343 and the depreciation rate of capital to
k
= 0:055.15 To select the parameter governing the elasticity of substitution, , we use
standard estimates of this elasticity that suggest a value of 1.5 – see Katz and Murphy
(1992) and Heckman, Lochner and Taber (1998). This dictates = 1=3. To calibrate the
share parameter , we force the model to reproduce the aggregate skill premium in the data,
de…ned as per-worker earnings of workers in the skilled category to per-worker earnings of
workers in the unskilled category. For this statistic, we target a value of 1.8.16 Tables A11
14
We consider all individuals who are not in armed forces.
15
We calibrate the capital share and the depreciation rate using a notion of capital that includes …xed
private capital, land, inventories and consumer durables. For the period 1960-2000, the resulting capital to
output ratio averages 2.93 at the annual level. We estimate the capital share and the capital to output ratio
following the standard methodology; see Cooley and Prescott (1995). The data for capital and land are from
Bureau of Economic Analysis (Fixed Asset Account Tables) and Bureau of Labor Statistics (Multifactor
Productivity Program Data).
16
The empirical target for the skill premium is from our calculations using data from the 2005 American
Community Survey (ACS). We restrict the sample to the civilian adult population of both sexes, between
ages 25 and 54 who work full time, and exclude those who are unpaid workers or make less than half of the
minimum wage. Full time workers are de…ned as those who work at least 35 hours per week and 40 weeks
per year. We estimate a value tightly centered around 1.8, when we include self-employed individuals or not.
22
and A12 shows full set of parameters.
Table A11: Parameter Values - Idiosyncratic Shocks

Benchmark Calibration
Statistic Permanent Shocks Persistent Shocks
Variance Single Skilled Males 0.2980 0.0063
Variance Single Unskilled Males 0.2570 0.0036
Variance Single Skilled Females 0.2510 0.0019
Variance Single Unskilled Females 0.2440 0.0018
Variance Married Skilled Males 0.2520 0.0068
Variance Married Unskilled Males 0.2270 0.0038
Variance Married Skilled Females 0.2240 0.0040
Variance Married Unskilled Females 0.2500 0.0008
Covariance (male, female) 0.0580 0.0010
5 Benchmark Economy - Additional Outcomes

In this section of the Appendix we present two additional outcomes that are mentioned in
the paper. Figure A8 shows the variance of log household consumption in the data and
the model. The model does an excellent job matching the level of inequality in household
consumption at the start of the life cycle and the size of increase along the life cycle.
0.60
0.50
Variance of Log Consumption
0.40
0.30
0.20
25 30 35 40 45 50 55
Age
data model
Figure A8 - Variance of Log Household Consumption
Figure A9 shows what happens to the variance of log wages and the labor force partici-
pation of married females in the benchmark economy when we set all childcare costs to zero,
while keeping all other parameters constant. The children matter critically in determining
the levels of participation rates, and how inequality in wages and earnings evolve over the
life-cycle for married females. When childcare costs are set to zero, the participation rate
of unskilled married females is much higher/ Furthermore, without children, the variance of
log wages grows linearly along the life cycle for women, exactly as it does for men.
23
Labor Force Participation, Skilled Females, Married Variance of Log Wages, Skilled Females
0.85
0.40
0.35

0.80
0.30
0.75
0.25
0.70
0.20
25 30 35 40 45 50 55 25 30 35 40 45 50 55
Age Age
data model model, no children data model model, no children
Figure A9 - LFP (left); Var. of Log Wages (right), Married Skilled Females
Table A12: Parameter Values

Benchmark Calibration
Parameter Value Comments
Population Growth (n) 0.011 U.S. Data
Discount Factor ( ) 0.9829 Calibrated - matches K=Y
Labor Supply Elasticity ( ) 0.2 Literature estimates.
Disutility from work, (Bf ; Bm ) 82.15, 28.67 Calibrated
Preference Shock 4 1:9 See text –Matches variance log hours at age 40
Skill depreciation, females ( x ) 0.025, 0.059 Calibrated
Growth of skills ( ex ) - See text - CPS data
Distribution of utility costs (:jz) - See text - matches LFP by education
(Gamma Distribution) conditional on husband’s type
Loading Factor #x (tmin ) - See text –matches LFP by age of youngest child
Capital Share ( ) 0.343 Calibrated
Skilled Labor Share ( ) 0.5085 Calibrated
Substitution Elasticity ( ) 1=3 Literature estimates
Depreciation Rate ( k ) 0.055 Calibrated
Childcare costs for single females, - See text - matches expenditure by age,
dS (x; t) and skills.
Childcare costs for married females - See text - matches expenditure by age,
M
d (x; t) and skills.
Tax functions T M (I; k) and T S (I; k) - See Appendix - IRS Data
M
Transfer functions T R (I; k), - See text and Appendix
S S
T Rf (I; k) and T Rm (I; k) -
Payroll Tax Rate ( p ) 0.086 See Appendix
Social Security Incomes, - See Appendix - U.S. Census
pSm (z), pSf (x) and pM (x; z) -
Capital Income Tax Rate ( k ) 0.097 See Appendix - matches
corporate tax collections
24
6 Optimal NIT
Figure A10 displays how aggregate output, ex-ante welfare for all, and majority support
changes with the NIT transfers. When the transfer equals zero, the tax system is simply
a proportional tax with no transfers whatsoever, and output is about 3.2% higher than in
the benchmark case. As transfers increase, tax rates, welfare and popular support increase
as well, but changes in output relative to the benchmark case become gradually lower and
eventually become negative. Figure 10 shows that as the lump-sum transfer increases, both
welfare and support for the reform …rst sharply increase and then decline. For a transfer
level of about 6% of mean income, ex-ante welfare gains are negative and majority support
disappears. At this level, the tax rate required is not trivially higher than at the welfare-
maximizing level (about 23.8%). Output is about 1.8% lower than in the benchmark case.
NIT NIT
80
3
0
2
-1
-1
70
Welfare Gain/Loss, %
Welfare Gain/Loss, %
1
-2
-2
winners
output
60
-3
-3
0
50
-1
-4
-4
40
-5
-5
-2
0 2 4 6 0 2 4 6
Transfer Level, percentage of mean household income Transfer Level, percentage of mean household income
Welfare Gain/Loss Winners Welfare Gain/Loss Output
Figure A10 - Welfare Gains and Winners, NIT(left); Welfare Gains and Output, NIT (right)
7 Benchmark Economy with CE Pro…les

Figure A11 shows that a recalibrated version of the benchmark economy has no trouble
matching the age-inequality pro…les produced using a cohorts-e¤ects speci…cation. Table
A13 shows the other model outcomes. The parameter values we use for this alternative
benchmark are presented in Tables A17-19.

0.60
0.60
0.50
0.50
0.50
0.50
0.40
0.40

0.40
0.40
0.30
0.30
0.30
0.30
0.20
0.20
0.20
0.20
25 30 35 40 45 50 55 25 30 35 40 45 50 55 25 30 35 40 45 50 55 25 30 35 40 45 50 55
Age Age Age Age
Unskilled, Data Skilled,Data Unskilled, Data Skilled,Data Unskilled, Data Skilled,Data Unskilled, Data Skilled,Data
Unskilled, Model Skilled, Model Unskilled, Model Skilled, Model Unskilled, Model Skilled, Model Unskilled, Model Skilled, Model
Figure A11 - Variance of Log Wages, Model vs. Data, Males (left), Females (right), data with CE
25
Table A13: Model and Data (YE and CE Calibration)
Aggregates Data BM (YE) Data CE
Capital Output Ratio 2.9 2.9 - 2.9
Total Transfers (% of GDP) 2.3 2.3 - 2.4
Skill Premium 1.8 1.8 - 1.8
Unskilled 68.2 68.7 - 68.5
Skilled 77.4 77.7 - 78.8
Total 71.8 72.3 - 72.6
Variance log-wages (Married Males, age 54, S) 0.45 0.45 0.54 0.54
Variance log-wages (Married Males, age 54, U) 0.34 0.34 0.37 0.37
Variance log-wages (Married Females, age 54, S) 0.35 0.35 0.44 0.45
Variance log-wages (Married Females, age 54, U) 0.26 0.26 0.31 0.30
Variance log-hours (Married Females, age 40) 0.13 0.13 0.11 0.11
Correlation Between Wages of Spouses (age 25) 0.31 0.31 0.30 0.31
Correlation Between Wages of Spouses (age 40) 0.34 0.33 0.34 0.33
Variance log-consumption (Age 55 vs 25) 0.12 0.12 0.15 0.16
90-10 ratio 7.8 7.1 - 7.4
90-50 ratio 2.6 2.5 - 2.6
Share, bottom 10% 1.8 2.1 - 1.95
Share, bottom 20% 4.5 5.5 - 5.2
Share, bottom 40% 13.2 15.8 - 15.2
Note: Entries show model outcomes for benchmark economy and the case where moments created using
cohorts e¤ects, as discussed in Section 7 in the paper
8 Rethinking the Welfare State When Inequality is

Lower
In this section of the Appendix, we present calibration details for "the 1980" economy in the
paper. Recall that for the benchmark economy, we use life-cycle pro…les generated using the
CPS for the 1980-2019 period, complemented by cross-sectional facts from the 2008 American
Community Survey. For the 1980 economy, we estimate the life-cycle pro…les using CPS data
for the 1980-1994 period and use cross-sectional facts from the 1980 and 1990 US Census.
The 1980 economy di¤ers from the benchmark along three dimensions. First, only about
19% of females had a college degree in 1980, and this number more than doubled to nearly
39% in 2008. For males, the fraction with a college degree increased from 29% to 35%.
Substantial changes in marital sorting accompanied these changes; about 14% of married
households were of the skilled-skilled category in 1980, while the corresponding …gure in our
26
parameterization is nearly 27%. These facts for the 1980 economy are reported in Tables
A16 and A17 below.
Second, there has been a signi…cant increase in inequality. The skill premium was about
1.5 in 1980 and increased to 1.8 in our benchmark parameterization. The left panel in Figure
A12 shows the age-productivity pro…les for males for the 1980 and the benchmark economy.
In both …gures, hourly wages are normalized by mean hourly wages in the data for each
year, and relative wages of skilled are much higher in the benchmark. In Figure A13, we
report the variance of log wages by age, skill level, and marital status for the benchmark
(dash lines) and the 1980 (solid lines) speci…cations. The 1980 pro…les have a lower intercept
(lower inequality at age 25) and, particularly for women, imply a lower increase in inequality
along the life cycle.
Finally, the labor force participation of married females was lower in 1980. Table A16
shows the labor force participation of married females by their and their husbands’produc-
tivity. Compared to the numbers in Table A11, the labor force participation of married
females is about 4 (9) percentage points lower for couples composed of two skilled (un-
skilled) partners. Figure A14 show the labor force participation of married females by their
(left panel) and their children’s (right panel) age, calculated using data for the 1980-1994
and the 1980-2019 periods.
We capture the e¤ect of these changes on our results in two steps. First, we focus on
the role of inequality. To this end, we calibrate an alternative benchmark economy, where
as model inputs we use the 1980 demographics (Tables A14 and A15) and age-productivity
pro…les for males (Figure A13, left panel) and the associated wage growth rates, xj (Figure
A13, right panel) for females. We also target the life-cycle inequality pro…les estimated using
the 1980-1994 CPS data (Figure A15) and a skill premium of 1.5. We call this alternative the
1980 (I) case in the paper. Then, we also target the married female labor force participation
(Table A16 and Figure A15-left panel) and call it the 1980 (II) case.
In both exercises, all other model inputs and targets remain the same as in our benchmark
economy. Hence, these exercises should be interpreted as our benchmark economy with lower
levels of inequality and married female labor force participation rather than representations
of the 1980 US economy. The parameter values are reported in Tables A18-20.
Table A14: Distribution of Married Working Households by Type

1980
Females
Skilled 14.43 13.98
Note: Entries show the fraction of marriages out of the total married pool, by wife and husband educa-
tional categories. The data used is from the 1980 Census, ages 30-39. Entries add up to 100.
27
Table A15: Fraction of Agents by Type, Gender and Marital Status
1980
Males Females
All Married Singles All Married Singles
Unskilled 70.65 61.75 8.91 80.85 68.40 12.45
Skilled 29.35 24.50 4.85 19.15 15.44 3.71
Note: Entries show the fraction of individuals in each educational category, by marital status, constructed
under the assumption of a stationary population structure from the 1980 Census.
Table A16: Labor Force Participation of Married Females, 25-54

1980
Females
Skilled 58.15 69.80
Note: Each entry shows the labor force participation of married females ages 25 to 54, calculated from
the 1980 and 1990 Census (the average values are reported).
2.00
0.08
0.06
1.50
Normalized Wages
0.04
Wage Growth
0.02
1.00
0.00
-0.02
0.50
25 30 35 40 45 50 55 25 30 35 40 45 50 55 60 65
Age Age
Unskilled, 1980s Skilled, 1980s Unskilled, 1980s Skilled, 1980s

x
Figure A12 - Age-Labor Productivity Pro…les, Males (left), Female Wage Growth, j (right),
benchmark data vs. the 1980s
28
0.60
0.60
0.35
0.35
0.50
0.50
0.30
0.30


0.40
0.40
0.25
0.25
0.30
0.30
0.20
0.20
0.20
0.20
0.15
0.15
25 30 35 40 45 50 55 25 30 35 40 45 50 55 25 30 35 40 45 50 55 25 30 35 40 45 50 55
Age Age Age Age
Unskilled, 1980s Skilled, 1980s Unskilled, 1980s Skilled, 1980s Unskilled, 1980s Skilled, 1980s Unskilled, 1980s Skilled, 1980s
Unskilled Skilled Unskilled Skilled Unskilled Skilled Unskilled Skilled
Figure A13 - Variance of Log Wages, Males (left) and Females (right), benchmark data vs. the 1980s
0.90
0.90
0.80
0.80

0.70
0.70
0.60
0.60
0.50
0.50
25 30 35 40 45 50 55 0 5 10 15
Age Age of Youngest Child
Unskilled, 1980s Skilled, 1980s Unskilled, 1980s Skilled, 1980s

Figure A14 - LFP of Married Females (left panel), LFP by the age of youngest child (right panel),
benchmark data vs. the 1980s

0.60
0.60
0.35
0.35
0.50
0.50
0.30
0.30


0.40
0.40
0.25
0.25
0.30
0.30
0.20
0.20
0.20
0.20
0.15
0.15
25 30 35 40 45 50 55 25 30 35 40 45 50 55 25 30 35 40 45 50 55 25 30 35 40 45 50 55
Age Age Age Age
Unskilled, Data Skilled,Data Unskilled, Data Skilled,Data Unskilled, Data Skilled,Data Unskilled, Data Skilled,Data
Unskilled, Model Skilled, Model Unskilled, Model Skilled, Model Unskilled, Model Skilled, Model Unskilled, Model Skilled, Model
Figure A15 - Variance of Log Wages, Model vs. Data, Males (left), Females (right), the 1980 (I)
29
9 Other Parameterizations
In this section of the Appendix, we present the parameter values that are used for di¤erent
economies discussed in Section 7 of the paper (Tables A17-A19). Table A20 shows the model
outcomes.
Table A17: Parameter Values - Permanent Shocks

Di¤erent Cases
Statistic BM Cohort 1980 I 1980 II = 1:5
E¤ect (scale econ.)
Variances
Single Skilled Males 0.298 0.298 0.245 0.245 0.298
Single Unskilled Males 0.257 0.259 0.241 0.241 0.257
Single Skilled Females 0.251 0.24 0.204 0.204 0.251
Single Unskilled Females 0.244 0.242 0.207 0.207 0.244
Married Skilled Males 0.252 0.243 0.204 0.204 0.252
Married Unskilled Males 0.227 0.227 0.224 0.224 0.227
Married Skilled Females 0.224 0.224 0.192 0.192 0.224
Married Unskilled Females 0.250 0.228 0.220 0.220 0.250
Covariance (male, female) 0.058 0.059 0.043 0.039 0.053
Table A18: Parameter Values - Persistence Shocks

Di¤erent Models
Statistic BM Cohort E. 1980 I 1980 II = 1:5 = 1:5
Variances (scale econ.)
Single Skilled Males 0.0063 0.0079 0.00727 0.00727 0.00627 0.00627
Single Unskilled Males 0.0036 0.0042 0.00406 0.00406 0.00356 0.00356
Single Skilled Females 0.0019 0.0060 0.00150 0.00150 0.00190 0.00190
Single Unskilled Females 0.0018 0.0035 0.00195 0.00195 0.00175 0.00175
Married Skilled Males 0.0068 0.0101 0.00642 0.00642 0.00675 0.00675
Married Unskilled Males 0.0038 0.0050 0.00400 0.00400 0.00380 0.00380
Married Skilled Females 0.0040 0.0072 0.00270 0.00270 0.00400 0.00400
Married Unskilled Females 0.0008 0.0028 0.00130 0.00130 0.00080 0.00080
Covariance (male, female) 0.0010 0.0017 0.00172 0.00172 0.001 0.0010
30
Table A19: Parameter Values
Di¤erent Cases
Parameter Value Cohort E. 1980 I 1980 II = 1:5
(scale econ.)
Discount Factor ( ) 0.9829 0.9825 0.9830 0.9829 0.9976
Preference Shock 4 1.88 1.70 1.957 2.055 1.9
Skill depreciation, females
s 0.025 0.025 0.025 0.025 0.025
u 0.056 0.056 0.056 0.056 0.056
Skilled Labor Share ( ) 0.505 0.505 0.3745 0.3715 0.509
Table A20: Model and Data

Di¤erent Cases
Aggregates Data BM = 1:5
(scale econ.)
Capital Output Ratio 2.9 2.9 2.9
Total Transfers (% of GDP) 2.3 2.3 2.3
Skill Premium 1.8 1.8 1.8
Unskilled 68.2 68.7 68.1
Skilled 77.4 77.7 78
Total 71.8 72.3 72
Variance log-wages (Married Males, age 54, S) 0.45 0.45 0.45
Variance log-wages (Married Males, age 54, U) 0.34 0.34 0.34
Variance log-wages (Married Females, age 54, S) 0.35 0.35 0.35
Variance log-wages (Married Females, age 54, U) 0.26 0.26 0.27
Variance log-hours (Married Females, age 40) 0.13 0.13 0.13
Correlation Between Wages of Spouses (age 25) 0.31 0.31 0.29
Correlation Between Wages of Spouses (age 40) 0.34 0.33 0.31
Variance log-consumption (Age 55 vs 25) 0.12 0.12 0.10
90-10 ratio 7.8 7.1 6.5
90-50 ratio 2.6 2.5 2.4
Share, bottom 10% 1.8 2.1 2.3
Share, bottom 20% 4.5 5.5 5.9
Share, bottom 40% 13.2 15.8 16.8
31
10 Elimination of Transfers
In Tables A21 and A22 we present the aggregate and welfare e¤ects of eliminating di¤erent
transfers one at a time. The elimination of means-tested transfer programs or traditional
’welfare’programs, has the largest impact. This elimination leads to an increase in output
in the long run of 1.1% –nearly two thirds of the increase when all programs are eliminated.
Hours worked increase by 1.9% and participation rates of unskilled (all) married women
goes up by 4.0% (2.9%). The aggregate …ndings associated to the elimination of individual
programs have a counterpart in terms of welfare e¤ects. The elimination of traditional welfare
programs leads to an ex-ante welfare loss of 1.3%, with unskilled single females experiencing
the largest loss (-3.%). Nonetheless, there is a concomitant majority support as taxes as
reduced for the majority. Interestingly, the elimination of child-related transfers has the
second-largest welfare loss but without majority support for its elimination. This occurs as
its elimination impacts narrowly households with children.
Table A21: Eliminating Transfers (% changes relative to benchmark)

Eliminating Eliminating Eliminating Eliminating
All Transfers Welfare EITC Child-Related
Programs Program Programs
Output 1.7 1.1 0.4 0.1
Aggregate Hours 3.0 1.9 0.9 0.1
Hours per worker (All Females) 3.2 1.6 0.3 0.8
Hours per worker (All Males) 1.9 1.2 0.5 0.5
Unskilled 6.3 4.0 4.0 -1.9
Skilled 2.0 1.4 1.1 -0.6
Total 4.5 2.9 2.8 -1.4
Entries in the top panel show the e¤ects (percentage changes) across steady states on selected variables
driven by the elimination of di¤erent transfer programs, and all of them simultaneously (the entire ’welfare
state’).
32
Table A22: Eliminating Transfers - Welfare E¤ects (Newborns, %)
Eliminating Eliminating Eliminating Eliminating
All Transfers Welfare EITC Child-Related
Programs Program Programs
Single F
Unskilled -5.2 -3.0 -0.9 -0.6
Skilled -1.2 -1.0 -0.1 -0.1
Married
Unskilled, Unskilled -0.1 0.8 -0.0 -0.7
Unskilled (f), Skilled (m) 0.4 0.3 0.1 -0.0
Skilled, Skilled 1.7 1.3 0.3 0.1
Skilled (f), Unskilled (m) 0.6 0.5 0.2 -0.1
All Newborns -3.2 -1.3 -0.3 -0.8
Winning Households 60.7 66.5 81.0 47.5
Entries show the welfare e¤ects (consumption compensation) driven by the elimination of di¤erent trans-
fer programs, and all of them simultaneously (the entire ’welfare state’). The calculations report welfare
gains across steady states under the assumption that the rental rate of capital (and interest rate) is constant
across steady states.
References
[1] Cooley, Thomas F. and Edward C. Prescott. (1995): “Economic Growth and Business
Cycles” in Frontiers of Business Cycle Research, pages 1-18, Thomas F,. Cooley (ed.),
Princeton University Press, Princeton: New Jersey.
[2] Guner, Nezih, Kaygusuz, Remzi and Gustavo Ventura. (2014): "Income Taxation of
U.S. Households: Facts and Parametric Estimates," Review of Economic Dynamics, 17,
559-581.
[3] Guner, Nezih, Remzi Kaygusuz, and Gustavo Ventura. (2020): “Child-Related Transfers,
Household Labor Supply and Welfare." Review of Economic Studies, 87, 2290–2321.
[4] Heathcote, Jonathan, Perri, Fabrizio and Giovanni Violante. (2010): "Unequal we stand:
An empirical analysis of economic inequality in the United States, 1967–2006." Review
of Economic Dynamics, 13, 15-51.
[5] Heathcote, Jonathan, Storesletten, Kjetil and Gianluca Violante. (2010): “The Macroeco-
nomic Implications of Rising Wage Inequality in the United States,”Journal of Political
Economy, 118, 681-72.
[6] Heckman, James J. (1979): "Sample Selection Bias as a Speci…cation Error." Economet-
rica, 47,153-161.
[7] Heckman, James, Lance Lochner and Christopher Taber. (1998): “Explaining Rising
Wage Inequality: Explorations With A Dynamic General Equilibrium Model Of Labor
Earnings With Heterogeneous Agents," Review of Economic Dynamics, 1, 1-58.
33
[8] Katz, Lawrence and Kevin Murphy. (1992): “Changes in Relative Wages, 1963-1987:
Supply and Demand Factors," The Quarterly Journal of Economics, 107, 35-78.
[9] Huggett, Mark, Ventura, Gustao, and Yaron, Amir. (2011): "Sources of Lifetime Inequal-
ity." American Economic Review, 101, 2923–2954.
34

Rethinking The Welfare State August 2023

Uploaded by

Copyright:

Available Formats

Rethinking The Welfare State August 2023

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Rethinking The Welfare State August 2023

Uploaded by

Copyright:

Available Formats

Rethinking the Welfare State

Nezih Guner, Remzi Kaygusuz and Gustavo Ventura

Guner, CEMFI, Spain (email: nezih.guner@cem….es); Kaygusuz, Department of Economics,

Findings We conduct three sets of experiments. First, we consider the hypothetical

Single Males Married Males Single Females Married Females

Variance of Log Wages

Variance of Log Wages

Variance of Log Wages

Unskilled Skilled Unskilled Skilled Unskilled Skilled Unskilled Skilled

Figure 1 - Variance of Log Wages, Males (left) and Females (right)

Gender Gap Married Female LFP

Labor Force Participation

Unskilled Skilled Unskilled Skilled

6. Conditional on work, there is signi…cant variation in hours of work among married

3 The Economic Environment

with "Sm;z;j+1 N (0; 2

h0 = H(x; h; l; e) = exp [ln h + e

wz $(z; j) exp( Sm;z + S

3.2 Children and Childcare Costs

3.4 Taxes and Transfers

3.5 Decision Problem

VfS (a; h; e; SfS ; ; j) = max

Meanwhile, D stands for the childcare expenses incurred:

c + a0 = a(1 + r(1 k )) + EfS (x; h; ; ; l)(1 p)

+T RfS (I; 0; 0) T S (I; 0):

iii) Retired: if j JR , then there are no children and

c + a0 = a(1 + r(1 k )) + pSf (x) T S (ra; 0) + T RfS (ra; 0; 0):

e0 = e + (l) and l 0, a0 0 (with strict equality if j = J + 1):

V M (a; h; e; S M ; ; j) = 0max fU M (c; lf ; lm ; q; x; z; b; j) + E 0j V M (a0 ; h0 ; e0 ; S M ; 0

where I = E M (x; z; h; ; ; lm ; lf ; j) + ra; and lm 0, lf 0; and a0 0:

Endowments For males, following the procedure described in Section 2, we construct

Transfers In the Appendix, we provide a description of the various means-tested pro-

0 50000 100000 150000 200000

Figure 4 - Total Transfers in the Benchmark Economy

Remaining Parameters We select the remaining parameters to match jointly several

5 The Benchmark Economy

Life-cycle statistics Our model environment is consistent with a host of observations

Variance of Log Wages

Variance of Log Wages

Variance of Log Wages

Labor Force Particiption

5.1 How Valuable is the Welfare State?

6 Rethinking the Welfare State

6.1 A Universal Basic Income

6.2 A Negative Income Tax

Table 3: Aggregate and Inequality Findings

Proportional Tax Rate (%) - 19.8 19.8

7.1 Rethinking the Welfare State When Inequality is Lower

Table 4: Welfare E¤ects

All Newborns -1.3 0.2 0.7

Including Transitional Dynamics

7.3 Funding Transfers with Higher Progressivity

Table 5: Findings in Perspective (% changes relative to benchmark)

Output -0.6 0.7 0.6 0.7 -3.8

1 Transfers to Households in the United States

1.1 The Earned Income Tax Credit

1.2 Child-Related Transfers

1.3 Welfare System

Supplemental Nutrition Assistance Program The SNAP is a federal program that