Journal Pre-proof

Cost of Policy Choices: A Microsimulation Analysis of the Impact on

Family Welfare of Unemployment and Price Changes

Julie L. Hotchkiss , Robert E. Moore , Fernando Rios-Avila

PII: S0164-0704(19)30051-5
DOI: https://doi.org/10.1016/j.jmacro.2019.103167
Reference: JMACRO 103167

To appear in: Journal of Macroeconomics

Received date: 6 February 2019

Revised date: 29 October 2019
Accepted date: 30 October 2019

Please cite this article as: Julie L. Hotchkiss , Robert E. Moore , Fernando Rios-Avila , Cost of Policy
Choices: A Microsimulation Analysis of the Impact on Family Welfare of Unemployment and Price
Changes, Journal of Macroeconomics (2019), doi: https://doi.org/10.1016/j.jmacro.2019.103167

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.

© 2019 Published by Elsevier Inc.

Cost of Policy Choices: A Microsimulation Analysis of the Impact
on Family Welfare of Unemployment and Price Changes
Julie L. Hotchkiss*
Federal Reserve Bank of Atlanta
and Georgia State University
[email protected]
404-498-8198

Robert E. Moore
Georgia State University
[email protected]
404-413-0056

Fernando Rios-Avila
Levy Economics Institute of Bard College
[email protected]
845-758-7719

October 28, 2019

Key Words: Family welfare, joint labor supply, microsimulation, dual mandate,
monetary policy

JEL Codes: I30 Welfare, Well-being, and Poverty

E52 Monetary Policy
J22 Time Allocation and Labor Supply
D19 Household Behavior and Family Economics

*
Corresponding author. The views expressed here are not necessarily those of the
Federal Reserve Bank of Atlanta, or the Federal Reserve System. Research
assistance from Augustine Denteh is much appreciated, and thanks is extended to
Julie Cullen and Raj Chetty for making programs used as a framework for
estimating UI benefits available and to David Altig, James Alm, Kelly Chen,
Andrew Friedson, Mei Dong, Tim Dunne, Nicardo McInnis, Kelsey O‘Connor,
Andrew Oswald, Luigi Pistaferri, John Robertson, Šarlota Smutná, Ling Sun,
Stephen Vogel, and Randall Wright for helpful comments and suggestions, and to
referees for their thoughtful suggestions. The authors declare that they have no
relevant or material financial interests or conflicts that relate to the research in this
paper.
 Abstract
This paper calculates the welfare cost to families of an unemployment shock.

Using U.S. data, we find an average annualized expected dollar equivalent

welfare loss of $1,156 when the unemployment rate rises by one-percentage

point. Relative to single families, the welfare loss is greater for married families

and increases with education. We also estimate that a loss in purchasing power of

1.8 percent generates the same amount of welfare loss as a one percentage point

rise in the unemployment rate. Additionally, the magnitude of the shock to

purchasing power that a family is willing to endure to avoid a one percentage

point increases in the aggregate unemployment rate rises with income. The results

in this paper informs policy makers about the distributional implications of

decisions likely to affect labor markets.

Cost of Policy Choices: A Microsimulation Analysis of the Impact
on Family Welfare of Unemployment and Price Changes

1. Introduction and Background

The purpose of this paper is to evaluate the welfare cost of a shock to

unemployment for families with different characteristics. We apply the

microsimulation methodology to estimate parameters of a labor supply model

within the context of a family utility framework for married couple households,

and within the context of a unitary utility framework for single households.

Estimated parameters from the utility model are used to simulate the expected

welfare loss from a rise in the aggregate unemployment rate, accounting for the

negative income shocks and changes in non-market time, with the recognition that

each person's probability of unemployment is impacted differently by a softening

of the labor market.1

Previous studies have explored the costs of unemployment almost

exclusively through a macroeconomic lens. Okun's Law (Okun 1962) is often

used to describe the loss in output that is generated from an additional one-

percentage point rise in the unemployment rate. Gordon, Nordhaus, and Poole

(1973) detail the deficiencies of Okun's Law (alone) for measuring the welfare

effects of a rise in the unemployment rate because the relationship does not

1
Non-market time is a combination of time spent on leisure, household
production and other activities outside paid labor.

-1-
account for the value of non-market activity. And rather than explore the cost of a

specific shock to unemployment, some focus more on the welfare costs of

economic volatility over time (e.g. Lucas 1991; Krusell and Smith 1999).2

An exception to the macroeconomic approach to measuring the welfare

costs of unemployment is found in Hurd (1980). Hurd uses estimated individual

labor supply elasticities (or, rather, the slope of the labor supply function) to

calculate the payment required to make a person indifferent between working the

desired hours at a prevailing wage rate, or being forced to work fewer hours than

desired because of unemployment. This payment is interpreted as the cost of

unemployment. Our methodology employs a similar concept in that we estimate

the welfare loss of deviating from desired hours, but we estimate utility function

parameters in order to calculate actual loss in welfare (i.e., utility) as opposed to

just the loss in income that would come from unemployment. Among other

things, this allows us to account for any potential welfare gain from an increase in

non-market activity that comes with non-employment. The methodology also

allows a comparison of welfare loss across families of different characteristics,

irrespective of the actual utility level of those families (either independently or

relative to one another).

2
Barlevy (2005) provides a good overview of the literature concerned with the
welfare cost of volatility.

-2-
 DiTella, MacCulloch, and Oswald (2001) also offer an estimate of the

utility-constant cost of unemployment and provide a segue to the second part of

the analysis in this paper. They assess the relative importance of high

unemployment vs. high inflation in explaining variations in demographic-neutral

aggregate levels of satisfaction across countries and time. They find that

unemployment reduces overall satisfaction more than inflation. They motivate

their analysis by stating that, "... reducing inflation is often costly, in terms of

extra unemployment...," (DiTella, MacCulloch, and Oswald 2001, 335). This

trade-off is also acknowledged by Gordon, Nordhaus, and Poole (1973) as a

motivation for undertaking their assessment of the welfare cost of higher

unemployment. However, they note that their assessment will not take account of,

"...the benefits associated with the lower inflation rate made possible by higher

unemployment," (p. 135).3 De Neve et al. (2017) also find a significant positive

relationship between changes in macroeconomic conditions and individuals'

assessments of well-being with the added revelation that negative shocks have a

more dramatic effect on well-being than positive shocks.4

3
In spite of this implied negative relationship between unemployment and
inflation, Berentsen, Menzio, and Wright (2011) identify a positive relationship
between unemployment and inflation in very low frequency data (the long run).
4
A related literature is concerned with macroeconomic levels and subjective well-
being. For example, see Proto and Rustichini (2013) and Stevenson and Wolfers
(2013).

-3-
 The potential trade-off between unemployment and inflation suggested by

these papers is of particular interest to U.S. Federal Reserve monetary policy

makers whose actions are guided by what is known as the "Dual Mandate" of full

employment and stable prices, which is spelled out in Section 2A of the Federal

Reserve Act:

―The Board of Governors of the Federal Reserve System and the Federal Open
Market Committee shall maintain long run growth of the monetary and credit
aggregates commensurate with the economy‘s long run potential to increase
production, so as to promote effectively the goals of maximum employment,
stable prices, and moderate long-term interest rates.‖ 5

While we do not model inflation in this paper, the second part of the analysis

estimates the size of an exogenous shock to purchasing power that would generate

the same welfare loss as a one percentage-point shock to unemployment. Since

the only consumption price in the model is the numeraire price of consumption,

we simulate a loss in purchasing power by adjusting the value of the other

components of the model that enter in real dollars -- wages and non-labor income.

We are then able to say something about how the individual family views the

trade-off between rising unemployment and decline in purchasing power.

The distributional implications of monetary policy has a fairly long

history. Krusell and Smith (1999) and Krusell et al. (2009) demonstrate that

Lucas' (1991) very small estimates of the cost of business cycle volatility doesn't

hold at the lowest points of the income distribution (also see Mukoyama and

5
See http://www.federalreserve.gov/aboutthefed/fract.htm.

-4-
Şahin 2006). And, while there is general agreement that monetary policy is not

responsible for the secular increasing trend in inequality since the early 1980s, it

is also agreed that monetary policy has a differential impact across the income

distribution and, hence, has a role to play in cyclical inequality (for example, see

Bernanke 2015; Nakajima 2015; Amaral 2017; Carpenter and Rodgers III 2004;

Coibion et al. 2017). The analysis in this paper finds differential welfare

implications for changing labor market environments by education and marital

status. In considering the welfare costs of unemployment and loss in purchasing

power, we do not suggest that the FOMC thinks of unemployment or inflation as

policy levers, but, rather, that these are economic outcomes that can be influenced

by policy choices. If this were not the case, then the Dual Mandate would be

pointless.

We find that the annualized expected welfare loss generated by a one-

percentage point shock to the unemployment rate is equivalent to $1,156, on

average across all families. And even though the probability of job loss is less for

those with higher education, their potential income loss is greater, making the

expected welfare loss for those with higher education greater than for those with

relatively lower education. In addition, married families‘ expected loss is greater

than that of single families (both in levels and as a share of total annual income).

This higher expected loss for married families translates into a higher equivalent

loss in purchasing power for married families than for single families; married

-5-
families are willing to tolerate greater loss to purchasing power to avoid

unemployment as compared to single families are.

2. Methodology

Microsimulation is a popular methodology often applied to assess the

impact of a specific policy on welfare (for example, see Fiorio 2008; Blundell et

al. 2000; Bahl et al. 1993; Blundell 1992; Gustman 1983). Here, rather than

evaluate a specific policy, we simulate the impact of the economic consequences

of any policy that is expected to negatively affect the labor market. The main

advantage to the theoretical framework we employ for this exercise is that it is

constructed from a standard joint (unitary for singles) family utility model. For

married couples, labor supply is jointly estimated. The utility function does not

include unemployment as a direct input in the optimization problem. However,

changes in unemployment and purchasing power can be brought to bear on the

welfare outcome by simulating the impact these environmental changes have on

behavior and family utility. 6

2.1. Family Utility Framework

The model described in this section nests the simpler case of single

households. Empirically, the single family version of the model implies

6
While economists often consider economic conditions (such as unemployment
or inflation) as an outcome of economic processes (see Hall 1981, 432), they are
certainly exogenous to an individual family's decision making process.

-6-
constraining hours and wages of the second household member to zero, as well as

constraining all utility parameters concerning the second member to be zero.

Family labor supply decisions are modeled in a neoclassical joint utility

framework often referred to as the "unitary" model. This model can be thought of

as a reduced-form specification of family decision-making. The model yields a

clear-cut expression of family welfare that allows for cross wage effects on each

member's labor supply decision. Assumptions of the unitary model are often

rejected in favor of a bargaining structure, or, more generally, the collective

model, for modeling intra-familial decisions making (for example, see Apps and

Rees 2009; McElroy 1990). However, a collective model framework provides no

concept of measurable household welfare, which is what we are after in this

analysis. What matters from the perspective of this article is how a policy

outcome impacts a family's welfare, providing less emphasis on the implications

in terms of decision-making structure within the household. Additionally, there is

evidence that the choice of structure for household decision making has very little

implication for conclusions in microsimulation exercises (see Moreau and

Bargain 2005). Further, Blundell et al. (2007) find that both collective and unitary

models are consistent with their household labor supply model estimated in the

U.K. We do not argue here that the unitary model is generally "better" than the

collective model, but rather that it is more appropriate for the research questions

in this article. The question posed in this paper requires differentiability of the

-7-
utility function in order to make use of the indirect utility function to draw

conclusions about changes in family welfare.7

Within the framework of the neoclassical family labor supply model, a

family maximizes a utility function that represents household welfare. Assuming,

for simplicity, that there are only two working members of the household

(husband and wife), the family chooses levels of non-market time (e.g., leisure,

household production) for each member and a joint consumption level in order to

solve the following problem:

( )
( )

. (1)

Define T as total time available for an individual; will be referred to

as the husband's non-market time, and will be referred to as the

wife's non-market time; is the labor supply of the husband; is the labor

supply of the wife; C is total money income (or consumption with price equal to

one); and are the husband's and wife's after-tax market wage, respectively;

and Y is non-labor income. and correspond to all uses of non-market time,

including home production activities. Since we are not concerned about the

distribution of different types of non-market time within the household in order to

7
Also see Browning, Chiappori, and Lechene (2006), who show that the unitary
model, unlike the collective model, is well behaved and satisfies the Slutsky
condition.

-8-
assess total family welfare, distinguishing leisure from household production is

not necessary. 8 In addition, the model does not distinguish between

unemployment and non-participation; both states are included in the non-

employment status. The implications of this are discussed later in section 2.3.

The solution to the maximization problem in equation (1) can be

expressed in terms of the indirect utility function, which is solely a function of the

wages of the husband and wife and non-labor income of the family:

( ) *, ( )- , ( )-

, ( ) ( ) -+ , (2)

where ( ) and ( ) correspond to the optimal labor supply

equations (desired hours) for the husband and wife, respectively. By totally

differentiating the indirect utility function, we can simulate the change in welfare

that results from changes in optimal hours of work and consumption in response

to changes in wages and non-labor income (also see Apps and Rees 2009, 263):

, (3)

8
Apps and Rees (2009) are highly critical of family utility models that do not
include measures of household production, but they acknowledge that not much
can be done without the availability of richer data (p. 108). The model presented
here uses an aggregated measure that includes household production, leisure as
well as other activities outside market job. Additionally, we do include the
number and age of children as determinants of labor supply decisions since the
presence of children may affect the comparative advantage between husbands and
wives in non-market work.

-9-
where and are the family's marginal utility of the husband's and wife's non-

market time, respectively, and is the family's marginal utility of consumption.

It is this equation that gives us the change in family welfare that will result from a

shock to unemployment or a shock to prices. It is clear from equation (3) that the

change in welfare not only depends on the individual labor supply responses, but

also on the family's marginal evaluation of a change in non-market time and

income.

2.2. Estimation of Utility Function Parameters and Labor Supply Elasticities

Simulating the impact on family welfare of higher unemployment and

shock to purchasing power requires the estimation of labor supply elasticities of

each family member with respect to changes in their own and each other's (in the

case of married-couple families) wages, elasticities with respect to non-labor

family income, as well as the changes in the probability of employment (extensive

margin elasticities); i.e., the probability of being at an interior solution on the

budget constraint. There are many divergent empirical issues raised in the

literature related to estimating labor supply elasticities. While the focus of this

paper is on the simulation exercise itself, the simulation does require labor supply

elasticities and it is, therefore, worthwhile to address some of the empirical issues.

Most of these issues, including the potential for endogeneity of wages and non-

labor income, are addressed in detail in online Appendix A. The goal here is to

produce reasonable labor supply elasticities; we deem them to be reasonable if

- 10 -
they are consistent with the literature. Toward that end, the methodology adopted

takes the simplest approach possible while maintaining basic theoretical and

empirical integrity. We also illustrate that the estimated labor supply elasticities

are well within the range of the existing literature, which contains significant

variation in modeling assumptions.

The requirement of simplicity here primarily derives from the goal of

quantifying the family-level utility changes. In order to obtain estimates of the

pieces of the change in utility in equation (3), a specific functional form of utility

must be specified. Following others (e.g., Hotchkiss, Moore, and Rios-Avila

2012; Hotchkiss, Kassis, and Moore 1997; Heim 2009; Ransom 1987), we

estimate a quadratic form of the utility function:

( ) ( ) ( ⁄ ) , (4)

where Z is a vector with elements , , and

;  is a vector of parameters and is a symmetric matrix of

parameters. This functional form has the advantage of being a flexible functional

form in the sense that it can be thought of as a second order approximation to an

arbitrary utility function. Furthermore, it is well-behaved when Amemiya's (1974)

internal consistency conditions are met (this is established in Appendix A). In

addition, it is possible to produce analytical closed-form solutions for both the

husband's and wife's labor supply functions. Obtaining the first order conditions

of this unconstrained maximization problem results in a system of equations

- 11 -
linear in :

=0 (5)

=0 (6)

This system can be solved simultaneously, and the desired hours become

( ) and ( ), which represent the desired number of

hours the members of a household would like to work, given the parameters that

define their household utility function, wages, and non-labor income. Details of

this derivation are reported in online Appendix B.

Observed hours ( ̃ ), however, might differ from the optimum hours due to

stochastic errors, such that:

̃ 2

̃ 2 , (7)

where we assume that ( ) follows a bivariate Normal distribution with mean

0 and covariance matrix ∑. This model can be thought of as a simultaneous Tobit

model, with working hours censored at zero, where we have four kinds of

families: those where both husband and wife work, those where only one of the

spouses works (two cases), and those where neither of them work. (Of course, for

singles, this simplifies to two cases -- the individual working or not working.)

Allowing for hours adjustment along the extensive margin for the wife when

- 12 -
assessing labor supply responses to wage changes have been found to make a

significant difference when assessing total labor supply response (for example,

see Heim 2009; Eissa, Kleven, and Kreiner 2008), however, extensive margin

hours adjustments appear to be unimportant for men (for example, see Heim

2009; Blundell et al. 1988). Considering the simulation of possible unemployment

for both men and women, allowing for husbands with zero hours of work is

important, so they will be included in the analysis.

Allowing for the presence of non-workers raises one empirical issue

identified by Keane (2011) that must be addressed: market wages are not

observed for individuals who do not work. To obtain estimates of those wages, we

take the standard approach in the literature of estimating a selectivity-corrected

wage equation (Heckman 1974) using regressors observable for both working and

non-working individuals.9 The resulting parameter estimates are then used to

predict wages for non-working men and women based on their observable

characteristics.

The maximum likelihood function corresponding to the joint labor supply

optimization problem can be written as follows:

9
For purposes of identification, the Heckman selection equation uses non-labor
income, number of children in the household, and spouse education (for married
households) as exclusion restriction variables.

- 13 -
 ( )
̃ ̃
∏ 6( ) 4 57

( )
̃ (̃ )
6 4 58 4 597
√

( ) ( )
̃ (̃ )
[ . /{ ( )}] . / , (8)
√

where and correspond to the probability density and cumulative distribution

functions of a univariate normal distribution, and and represent the

probability density and cumulative distribution functions of the bivariate normal

distribution. For singles, this likelihood function reduces to the univariate case.

Also, H=1 if the husband is working and W=1 if the wife is working (0

otherwise), (i=1,2) represents the standard deviations of ( ) and is the

correlation between the stochastic errors.

Obtaining reasonable estimates of labor supply elasticities is essential in

order to obtain credible estimates of the change in utility through the simulation

exercise described below. Issues that are well know in the literature relating to the

estimation of labor supply elasticities and the implications of those issues to the

problem at hand are addressed in detail in online Appendix A.

With the expectation of heterogeneity in preferences across families,

particularly of different age, education, and income levels (see Keane and Wasi

2016; and Deaton 2018), we estimate different sets of parameters for families

- 14 -
based on the husband's education level for married couples, and head of the

household's education for single families. In addition, we estimate different sets of

parameters for male and female singles. In other words, we estimate five sets of

parameters for married families (full sample; and four levels of husband's

education) and 10 sets of parameters for singles (full sample for men and women

separately; then for each education level separately for men and women). 10

2.3. Expected Welfare Loss from a Shock to Aggregate Unemployment

We simulate the impact of a rise in the unemployment rate as an

exogenous shock to the stochastic errors in equation (7). If, for example, an

employed husband loses his job, then . This also implies that the

estimated welfare impact of unemployment is, by construction, zero if neither of

the spouses is working.

The probability of each family member being hit by job loss is a function

of his or her demographic characteristics (gender, race, age, and education), as

well as time and location (details provided below). If the marginal effect on the

probability that the husband loses his job when the aggregate unemployment rate

rises by one-percentage point is and the marginal effect on the wife's

10
There are many other dimensions across which utility function parameters
could vary. We expect that differences across marital status and education/income
would be most pronounced, however additional heterogeneity (across age, race,
and number of children) is allowed for through the and found in equation
B1; estimates of their components are reported in Appendix E, Tables E1 and E2.

- 15 -
probability of losing her job is , then the expected change (loss) in family

welfare ( from equation 3) due to a positive probability of job loss is given by:

{ | } ( )( ) , -

+ ( ) , -

+( ) , -

+ , - (9)

The first term on the right hand side of equation (9) is the expected change

in utility if neither the husband nor the wife loses their jobs. The second term is

the expected change in utility from only the husband losing his job. The third term

is the expected change in utility from only the wife losing her job. And the last

term in this expression is the expected change in utility of both persons losing

their jobs. For singles, this expected utility reduces to just two terms

corresponding to the increased probability that the individual becomes non-

employed and one minus that probability.

The change in aggregate unemployment is assumed to be strictly an

exogenous shock and does not play a role when the family members choose the

optimal number of hours to work. And, except for being related through

characteristics a husband and wife might have in common (such as age, race, state

of residence, etc.), the marginal effects of job loss for husband and wife, in

- 16 -
married-couple households, are otherwise independent of each other. 11

When a family member loses his/her job, the family loses his and/or her

earnings, but that earnings loss may be offset somewhat by receipt of

Unemployment Insurance. Details of how we estimate the weekly benefit

allowance (wba), eligibility, and expected take-up rate ( ) are provided in online

Appendix C. 12 The fact that take-up rates are below 100 percent reflects the

choice of some individuals who lose their jobs to exit the labor force, rather than

remain unemployed.

The family may also be able to offset earnings loss through previous

savings. However, based on our calculations using the Survey of Consumer

Finances (SCF) (available upon request), it is unclear how the presence of savings

would differentially impact the estimations of expected welfare loss across

families from a rise in the unemployment rate. For example, in 2016, 48.9% of

households had zero liquid savings and 27.9% of those with savings had $6,000

or less in savings. Additionally, there is no consistent variation across income

deciles in households' average total liquid savings as a share of the average total

income (or as a share of average earnings). So, even though wealthier households

are more likely to save (i.e., more likely to spend less than they earn at any point

11
Additionally, market wages are assumed to be sticky (e.g., see Kahn 1997),
therefore assumed to not be a function of unemployment in this static framework.
12
For simplicity, we assume that all other sources of non-labor income are not
affected by the shock of unemployment.

- 17 -
in time, as also seen in the SCF), the ability of a typical high-income family to

replace lost earnings with savings (and maintain their usual level of consumption)

does not appear to be significantly different than that of a typical low-income

family. 13

The marginal impact of a change in the unemployment rate on the

probability of job loss is obtained by estimating the probability of non-

employment as a function of the aggregate unemployment rate.14 Each person is

assigned to one of 64 specific demographic groups (based on two gender, two

race, four age, and four education classifications). The impact of a rise in the

state/year aggregate unemployment rate on the probability of non-employment for

a member of that group is determined by 64 separate time-series probit

estimations using observations from the March supplement of the Current

Population Survey from the time period 2003-2013 with year and state fixed-

effects.15 We choose a 10-year period to average the marginal effects across the

most recent business cycle prior to the years of analysis. For example, the

13
Whereas higher-income families are more likely to have any savings at all to be
able to help smooth consumption, Aaronson et al. (2019) suggest that lower-
income families rely on credit to help smooth consumption in the event of a job
loss. Of course, the longer term consequences of depleting savings or exhausting
credit in the event of job loss is beyond the scope of this paper.
14
Details of the estimation procedure and a sample of estimated marginal effects
are provided in online Appendix D. An increase in non-employment (either from
unemployment or out of the labor force) will necessarily reflect job loss.
15
This procedure is similar to that employed by Gramlich (1974) in his
assessment of the distributional consequences of unemployment.

- 18 -
smallest marginal effect of a one-percentage point increase in the aggregate

unemployment rate was estimated to be a 0.056 percentage point decline in

employment for white women, between 35 and 44 years old, with at least a

college degree. Therefore, for a woman with these characteristics is set equal

to 0.00056. The largest marginal impact was estimated to be a 2.4 percentage

point employment decline for white men, between the ages of 18 and 34, with less

than a high school degree. Therefore, for a man with this set of characteristics

is set equal to 0.024. Given a set of estimated utility function parameters, and

estimated probabilities of job loss, the family-specific impact on expected utility

of a one-percentage point rise in the unemployment rate is given by equation (9).

The model does not explicitly depend on the labor market environment

(i.e., the prevailing aggregate unemployment rate) at the time of optimization. The

model specification assumes that whatever non-employment exists is optimal (or,

within a random error term of optimal). This means that if some of the observed

non-employment is technically unemployment, it is by choice – the person‘s

market/offered wage is less than his/her reservation wage. This optimization can

be thought of as taking place in the aggregate at the natural rate of unemployment.

We estimate utility function parameters using data from 2015-2016, a period of

time which most sources consider the economy to be at or near the natural rate of

unemployment (for example, see Federal Reserve Bank of Philadelphia 2017).

- 19 -
Therefore, this time period provides an environment in which we can interpret

observed non-employment behavior as near optimal.

3. Data

The Current Population Survey (CPS) is administered by the U.S. Bureau

of Labor Statistics each month to roughly 60,000 households.16 The survey has a

limited longitudinal aspect in that households are interviewed for four consecutive

months, not interviewed for eight months, then interviewed again for four months.

Households, families, and individuals can be matched across these survey months

if they remain in the same physical location. In survey months four and eight, the

household is said to be in the "outgoing rotation" group and members of the

household are asked more detailed questions about their labor market experience,

such as wages and hours of work.

We make use of the CPS outgoing rotation groups in March, April, May,

and June from 2015 and 2016 in order to construct the samples for which the

family labor supply model is estimated. We combine as many months as possible

across two years in order to construct the largest data set possible to meet the

demands of the challenging estimation problem. Detailed non-labor income is

obtained by matching each family to their March supplement survey, which is the

16
We obtained the CPS data set from IPUMS. See Flood et al. (2015).

- 20 -
month in which this information is collected. Households that couldn‘t be

matched to the March data are excluded from the analysis.

We restrict the sample further for two reasons. The first is for structural

reasons to make the observations conform better to the theoretical model. These

restrictions involve limiting the sample to households with members between 18-

64 years of age and excluding households with unmarried same- or opposite sex

adults/partners or children older than 18 years old. It is unclear in these

households how to assign the "husband" and "wife" labels and potential additional

adult labor supply is not accounted for in the model. We also exclude households

in which the main activity of both members is being a student, being retired, or

self-employment. We expect that those younger than 18, older than 64, students

and retired individuals have additional constraints on their optimization problem

not considered here. In addition, it is difficult to estimate market hourly earnings

(wage) for someone who is self-employed. Given the nature of their activities, in

a short period of time, reported earnings can be negative, even if, in the long term,

the market value of a self-employed worker's time would be positive.

Because the simultaneous estimation of nonlinear labor supply functions is

challenging, we also "trim" the data in various ways to eliminate outliers that

cause difficulties in the estimation process. Less than five percent of the sample is

eliminated based on the following restrictions: non-positive after-tax weekly

household income, negative non-labor income, after-tax hourly wages greater

- 21 -
than $600 or less than $0.50, or an estimated marginal tax rate 75 percent or

higher or lower than -60 percent.

Information on the detailed sources of non-labor income, number of

children, and earnings available from the CPS is used to calculate the marginal tax

rate on earnings (wages) and the total tax liability (in any year of interest) using

the National Bureau of Economic Research (NBER) TaxSim tax calculator. The

calculator is more complete than we have information for from the CPS, so we

made assumptions for the missing values as recommended by TaxSim managers.

For example, there is no information in the CPS that would allow one to calculate

itemized deductions (mortgage payments, charitable contributions, etc.), so values

of zero are entered for the missing information. Although unlikely to affect tax

rates, this will likely lead to an over-estimate of taxes paid for higher income

individuals, since they would likely receive a higher deduction through

itemization.17

Online Appendix E contains the means for the full sample and for each

sub-sample based on education, for married and single families (respectively). We

have a total of 20,163 married families and 15,485 single families in our sample.

17
http:// www.nber.org/~ taxsim/; see also Feenberg and Coutts (1993). In
addition to the detailed income source information from the CPS data, we also
include information on property tax, CPS imputed capital gains and capital losses.
All married households are classified as if they were declaring taxes jointly and
the main earner is identified as that with the highest total earned income. The tax
simulation was implemented using the Stata taxsim interface.

- 22 -
Among married families, about 88 percent of husbands and nearly 70 percent of

wives are working (with both percentages increasing in husband's education).

Husbands work more hours (43) and earn a higher after-tax hourly wage ($21.33

after tax) than their wives, who work about 37 hours and earn $16.16 after tax.

Husbands are slightly older than wives, at 45 vs. 43 years of age. Wives are

slightly more educated than their husbands. The families have roughly $347 per

week in (virtual) non-labor income. Virtual non-labor income is what the non-

labor income for the family would be if the portion of the non-linear constraint

they are on were extended to the vertical axis. The average federal (state)

marginal tax rate across families is 20 percent (4 percent).

Women comprise 56 percent of the single-persons sample. On average,

women have slightly more education; are slightly younger than the men; work

fewer hours (39 vs. 42 for single men); have about the same non-labor (virtual)

income; have a greater number of children; and earn lower wages. Roughly half

of all singles have never been married (46 percent of women and 54 percent of

men).

4. Results

4.1. Utility Function Parameter Estimates and Labor Supply Elasticities

Maximum likelihood estimates of the utility function parameters for both

married and single families are presented in online Appendix F, along with the

average labor supply elasticity and marginal utility estimates for married and

- 23 -
single families. For purposes of placing the estimated elasticities in context of the

literature, Figure 1 illustrates the intensive margin elasticities along with estimates

from previous studies. Note that own wage elasticities are averaged across

workers and non-workers. It is well known that varying assumptions can produce

a wide array of labor supply elasticities (see Mroz 1987); our estimates generally

fall within the range of those found in the literature.

- 24 -
Figure 1 Comparison of intensive margin elasticity estimates with the literature.
(a) Husband's (men's) elasticities

(b) Wife's (women's) elasticities

Notes: Sources of literature estimates are (Devereux 2004; Hotchkiss, Moore, and Rios-Avila 2012; Hotchkiss, Kassis, and Moore
1997; Heim 2009; Blau and Kahn 2007; Triest 1990; Pencavel 2002; Ransom 1987; Blundell and Macurdy 1999; Kumar 2009;
Bishop, Heim, and Mihaly 2009; Imai and Keane May2004; Chetty 2012; van Soest 1995). Also see Keane (2011) and McClelland
and Mok (2012) Many fewer sources provide estimates for participation elasticities, but ours fall within the literature bounds
(available upon request).

- 25 -
 Note that married women's own wage elasticities are higher than married

men's elasticities, indicating that women's labor supply is more responsive to

changes in their own wages. In addition, married women are more responsive to

changes in their own wages than are single women, who average an own-wage

elasticity very close to that of single men. The estimated negative cross-wage

elasticities (among married families) indicate that husbands and wives view

theirnon-market time as substitutes, which is consistent with the existing

literature. Cross wage elasticities for husbands and wives correspond to families

in which both members are working. Both men and women present the expected

negative income elasticity. The bottom line from these estimates is that the

simulation will be based on behavior reflected through labor supply elasticities

consistent with those estimated by others, using different data, empirical models,

and for different purposes. Online Appendix G provides a sensitivity analysis

(discussed more below) showing that our results are robust to variations in labor

supply elasticities.

4.2. Expected Welfare Loss from a Shock to Unemployment

By dividing the calculated expected loss in welfare from a one-percentage

point rise in the unemployment rate (in utils) by the family's marginal utility of

income/consumption ( ), we get a dollar value of that expected welfare loss.

Table 1 reports the annualized dollar value for the expected welfare loss from a

positive shock to unemployment for families of different types and education

- 26 -
levels (the loss as a share of total household income is also reported).

Workers at higher education levels earn higher wages, putting upward

pressure on the expected welfare loss from losing their job. For example, the

average annual income for families with at least a college degree is roughly twice

as large as for families with less than a high school degree (i.e., roughly $87,000

vs. $42,000). Therefore, families with higher education (higher earnings) have

much more to lose if they experience job loss. However, a higher education level

also means a lower probability of job loss, putting downward pressure on the

expected welfare loss from rising unemployment. For example, a one percentage

point rise in the unemployment rate increases the probability of job loss for

someone with less than a college degree by 1.23 percentage points, whereas the

marginal effect on someone with a college degree is only 0.54 percent points.

Table 1 Average welfare loss of a negative shock to employment by family type

and education.
Married Families Single Families
All education types $1,944 $131
[$1,735-$2,197] [$122-$141]
2.85% 0.29%
Less than high school $818 $7
[$587-$1,230] [$0-$57]
1.90% 0.03%
High school $1,501 $94
[$1,244-$2,001] [$78-$119]
2.47% 0.25%
Some college $1,769 $184
[$1,488-$2,169] [$158-$211]
2.55% 0.46%

- 27 -
 College or more $2,970 $1,039
[$2,382-$3,824] [$349-$4,165]
3.30% 2.04%
Single family type
married, spouse-not-present $153
[$128-$173]
0.32%
separated $125
[$106-$140]
0.31%
divorced $128
[$116-$138]
0.26%
widowed $75
[$60-$87]
0.16%
never married $139
[$125-$149]
0.32%
Note: Education refers to single head of household or husband education for
married families. Full sample estimates are used to report results for the full
sample and the education specific estimates are used to report results by education
group. 95 percent confidence intervals are in brackets; they were obtained through
bootstrapping with 199 repetitions. Percents reflect average across families of the
welfare loss as share of total family income.

Based on the results in Table 1, we see that, since the expected welfare

loss increases with education, the impact of the potential of losing higher wages

dominates the lower job loss probability. To better visualize the differences across

families, the estimates in Table 1 are plotted for both married and single families

(Figure 2, panel a). The largest difference in estimates across both married and

single families is that between the most and least educated, illustrating that the

- 28 -
loss of income is likely dominating the higher probability of employment loss.18

The average annualized expected welfare loss for the whole sample is

$1,156. This estimate of expected welfare loss is much lower than that found in

Hurd (1980), who estimates an individual welfare loss per unemployment spell of

about $7,000 (in 2012 dollars). Likely the main reason Hurd's estimate is so

much higher than ours is that we are estimating the expected welfare loss from

losing a job, rather than the actual cost of a specific job loss. Also, his model does

not allow for any positive utility gained from additional non-market time that

results from unemployment, nor the potential mitigating effects of unemployment

insurance. Additionally, Hurd does not allow for substitution of non-market time

between husband and wife, which will over-state the losses for men.

18
The very imprecise estimate of the welfare loss among college educated singles
could likely be coming from the estimated negative wage elasticity. Based on the
literature, this is more likely to be found among high-earning workers, suggesting
that the income effect from a wage change on hours dominates the substitution
effect.

- 29 -
Figure 2 Graphical representation of the annualized welfare loss and equivalent
loss in purchasing power arising from a one percentage point rise in the aggregate
unemployment rate (shown with 95 percent confidence intervals).
(a) Annualized dollar equivalent welfare loss

(b) Loss in purchasing power equivalent to welfare loss from a

one percentage-point rise in the unemployment rate

Note: See notes to Table 1.

- 30 -
 There is a significant difference between single and married families (in

both levels and as a percent of annual family income) -- the average expected

welfare loss overall is $1,944 among married families, whereas the average

annualized expected loss is only $131 among singles. To get some idea of where

this sizable difference might be coming from, we decompose the change in

welfare into each of its components: differences in non-employment probabilities,

change in hours for each family member, change in total consumption, and

changes in the marginal utilities (since we are moving families to a different point

on their indifference curve). The results of this decomposition are found in Table

2. Note that the decomposition is performed for the average married and average

single families, whereas the welfare losses reported in Table 1 reflect the average

loss across families (while the total loss estimates differ slightly, the relative

comparison is the same).

There are four possible employment outcomes for married families and

two possible employment outcomes for single families when the unemployment

rate increases. The lowest probability of job loss occurs for women in the average

married family, followed by the average single women, men in the average

married family, and then the average single man. Regardless of family type,

preferences behave as expected. When someone becomes non-employed (i.e., an

increase non-market time and a decrease in income), the marginal value of their

non-market time ( and ) declines and the marginal value of

- 31 -
consumption/income ( ) increases. One important difference between married

families and single families is that whenever the husband loses his job, the

family's marginal utility of his non-market time ( ) becomes negative -- there is

a tremendous loss in welfare from the husband not working.

Concluding that the welfare loss from a one percentage point rise in the

unemployment rate is greater for married families than for single families, and is

greater for the more educated, does not say anything about the welfare levels of

different family types or education levels. The losses estimated here from a rise in

the unemployment rate also do not imply that an analogous decline in

unemployment would generate a symmetric gain in welfare for families. In fact,

De Neve et al. (2017) find that subjective well-being is more sensitive to negative

economic conditions than to positive economic conditions. Further, the model in

this paper is not well-suited to assess employment gains since it is predicated on

the assumption of full employment.19

19
Welfare costs of a rise in the unemployment rate discussed here also do not take
into account the potential long term consequences of job loss on the mental and
physical health of those impacted and/or their children or on lifetime wealth (for
example, see Golberstein, Gonzales, and Meara 2016; Sullivan and Wachter
2009; Mathers and Schofield 1998; Krueger, Mitman, and Perri 2016; Gathmann
et al. 2018). Nor does our estimate of the expected welfare cost of rising
unemployment take into account any fear that families or individuals might have
of losing their job (as DiTella, MacCulloch, and Oswald 2001 claim their survey
of happiness does).

- 32 -
Table 2 Decomposition of the weekly welfare loss for the average married and average single family from a one percentage point
increase in the aggregate unemployment rate.
Average Married Family Average Single Man Average Single Woman
No job Both Husband Wife Single Single Single Single
loss husband loses job loses man man women women
and wife job does not loses job does not loses job
lose job TOTAL lose job TOTAL lose job TOTAL
98.20% 0.01% 1.08% 0.72% 98.84% 1.16% 99.22% 0.78%
0 -37.80 -37.80 0 0 -34.06 -- --
0 -25.92 0 -25.92 -- -- 0 -30.44
0 -967.63 -633.43 -334.20 0 -481.68 0 -384.00
5.602 -14.96 -14.14 4.78 132.51 86.02 -- --
4.221 0.56 3.59 1.19 -- -- 67.57 39.72
0.297 0.39 0.37 0.31 8.50 10.75 4.95 6.14

* + 0 -926.32 -769.93 -73.72 -8.92 0 -2,248.22 -26.08 0 -1,148.68 -8.96

$ equivalent 0 -$3,120 -$2,593 -$248 -$30 $0 -$209 -$3 $0 -$187 -$2
Annualized -$1,562 -$160 -$94
Note: Subscript "1" refers either to the husband, or single man; "2" refers to wife or single women. Notation corresponds to equations
(3) and (9). correspond to the marginal utilities of the husband's/single man's non-market time, wife's/single woman's non-
market time, and consumption, respectively. As an illustration of how the estimates in the table are used to construct * +,
consider the case of the average single man:
* + * + * + * + *
( ) ( )+ , where p is the probability of each outcome. This value is then divided by the marginal
utility of consumption ( ), under no job loss to get the weekly dollar equivalent of lost welfare.

- 33 -
4.3. Equivalent Welfare loss from a Shock to Purchasing Power

In order to illustrate how we simulate the loss in purchasing power needed

to generate the same expected welfare loss of a one percentage point rise in the

unemployment rate, the total derivatives in the indirect utility function from

equation (3) ( ) are expanded and the terms are rearranged to

isolate those reflecting changes in wages and non-labor income

( , ):

2 0 13

+2 0 13

+2 0 13 . (10)

The only consumption price in our model is that of the numeraire price of

consumption. Therefore, we can reflect a loss in purchasing power by changing

the other components that enter the model in real dollars. Equation (10) shows

how family welfare is affected by changes in wages and non-labor income,

directly, and also through each person's labor supply elasticities. Of course, there

are no cross-elasticities that enter the calculation in a single family's change in

utility.

If prices increase by i, one nominal dollar of income would only be able to

buy ( ) of any particular composite good. This implies that the value, or

purchasing power, of nominal wages and nominal non-labor income declines by

- 34 -
 , ( )-. Given that we are considering a one-time change in the level of

prices, we assume that nominal wages are sticky over the same time period, thus

there are no adjustments on wages or non-labor income over the horizon of

analysis (e.g., see Kahn 1997), hence there is a decline in purchasing power by the

same percent as the price increase.

Calculating the equivalent welfare cost from a loss in purchasing power,

then, amounts to finding the value of i that equates equation (9) and equation (10):

{ | } 2 0 13 . /

+2 0 13 . /

+2 0 13 . /. (11)

In other words, i is the percent increase in the consumption price level that

generates, for each family, the same expected change in utility as a one-

percentage point rise in the aggregate unemployment rate. This one-time price

level change will be able to tell us something about how the individual family

would experience the trade-off between a rise in unemployment and a loss in

purchasing power.

Table 3 presents the equivalent loss in purchasing power by marital status

and education. For the full sample, the average equivalent loss in purchasing

power is 1.82 percent (with a median of 0.83 percent). The results in Table 3 are

also illustrated in Figure 2, panel (b). The equivalent loss in purchasing power is

- 35 -
much lower among single families at 0.30 percent (0.21 percent at the median).

Since a one-percentage point rise in the unemployment rate is not as costly to

them, singles, if given a choice, would not willingly endure as large a loss in

purchasing power in order to avoid a rise in unemployment.

If we interpret this valuation of a shock to purchasing power as reflective

of a short-term reaction to unanticipated inflation, this result is consistent with

Burdett et al. (2016) who find that singles are much more likely to hold cash than

non-singles, which loses value more quickly with inflation than other assets; they

conclude that, "...inflation is a tax on being single" (p. 352).20 Although they

address the cost of inflation, and we simulate the cost from a loss in purchasing

power, the lower estimated cost from a loss in purchasing power among singles is

consistent with the conclusions of Burdett et al. (2016), who find that inflation

(which can be thought of as a loss in purchasing power if it affects prices and not

wages) is more costly for singles because, "...being single is cash intensive" (p.

337); also see (Dong, Sun, and Wright 2015). In addition, Burdett et al. find that

among the non-married, inflation is likely to be costliest to those who are

widowed; we find the same, as illustrated in Table 3.

20
Note that inflation can be thought of as a loss in purchasing power if it affects
prices and wages do not adjust (also see Aruoba, Davis, and Wright 2016; Dong,
Sun, and Wright 2015). Other research (Alm, Whittington, and Fletcher 2002) has
identified a tax on singles (relative to others with similar economic and
demographic characteristics) through the structure of the U.S. tax system.

- 36 -
Table 3 Average loss in purchasing power equivalent to the welfare loss from a
rise in the unemployment rate by one percentage point, estimated by family type
and education.
Married Families Single Families
All education types 2.98% 0.30%
[2.66%-3.44%] [0.27%-0.33%]
Less than high school 2.58% 0.11%
[1.29%-4.05%] [0%-0.39%]
High school 2.49% 0.23%
[2.01%-3.37%] [0.17%-0.31%]
Some college 2.57% 0.50%
[2.14%-3.54%] [0.41%-0.6%]
College or more 3.19% 2.28%
[2.54%-4.21%] [0.76%-34.15%]
Single Family Types
married, spouse-not-present 0.34%
[0.29%-0.39%]
separated 0.30%
[0.25%-0.35%]
divorced 0.28%
[0.25%-0.3%]
widowed 0.18%
[0.15%-0.2%]
never married 0.33%
[0.29%-0.37%]
Note: Education refers to single head of household or husband education for
married families. Full sample estimates are used to report results for the full
sample and the education specific estimates are used to report results by education
group. 95 percent confidence intervals are in brackets; they were obtained through
bootstrapping with 199 repetitions.

Another potential comparison for these results is the work by DiTella,

MacCulloch, and Oswald (2001). Their analysis across countries and time finds

that, "a 1-percentage-point increase in the unemployment rate equals the loss

brought about by an extra 1.66 percentage points of inflation" (p. 339). Again,

- 37 -
their analysis is quite different from the one presented here -- they estimate life

satisfaction as a function of unemployment and inflation across many countries

and time. And, although they do not provide the nuances seen here across

demographics, their estimated inflation generated loss of happiness (or, welfare)

equivalent to 1pp rise in the unemployment rate is within the range of the

equivalent loss in purchasing power presented in Table 3.

Online Appendix G contains the results from a sensitivity analysis for the

equivalent loss in purchasing power presented in Table 3. For the full sample of

married families, the alternate loss in purchasing power, using alternative

elasticities found in the literature, range from a low of 2.69 to a high of 3.95

percent. For single men and women, there is no measurable difference in the

estimates using alternative labor supply elasticities. The estimates presented here

for the expected welfare cost of unemployment and its equivalent shock to

purchasing power are clearly not being driven by differences found between our

labor supply elasticities and those in the rest of the literature.

4.4. Loss in Purchasing Power vs. Unemployment Shock Trade-off

There is a rich literature that estimates the cost of inflation in terms of how

much consumption one would be willing to give up to lower inflation. Estimates

- 38 -
range between 0.5 percent and 10 percent.21 The larger the estimate, the more

consumption one would be willing to give up to reduce inflation. There is a large

literature assessing the distributional effects of inflation (see Amaral 2017 for a

review) suggesting that since inflation is not as costly to wealthy individuals, they

would not be willing to give up as much consumption to avoid inflation. Or,

inversely, they would be willing to endure greater inflation to avoid higher

unemployment (potential loss to consumption).

Using results presented above, Figure 3 illustrates the shock to purchasing

power that is equivalent to the welfare loss from a rise in unemployment, that

welfare loss as a share of income (roughly, loss in consumption), and the ratio of

the two across the income distribution. The loss in purchasing power equivalent to

the welfare loss from a one percentage-point rise in the unemployment rate (panel

a) increases with income in the lower half of the income distribution, then

basically flattens out. The expected welfare loss from an increase in

unemployment, as a percent of income (panel b), also rises in the lower half of the

income distribution, but then declines as income continues to rise.

Since the ratio of the two (panel c) is increasing in income, higher income

families are willing to endure a greater loss in purchasing power to avoid a rise in

unemployment and lower income families are not willing to endure as much loss

21
See Lagos, Rocheteau, and Wright, (Forthcoming) for a review of estimates
from the literature.

- 39 -
Figure 3 Comparing welfare losses from a price shock vs. an unemployment shock across the income distribution, full sample averages.

(a) Welfare loss from a loss in purchasing power (b) Welfare loss from 1pp increase in the (c) Ratio of welfare loss from loss in
equivalent to the welfare loss from a rise in the aggregate unemployment rate (% of total purchasing power vs. welfare loss from
aggregate unemployment rate income) unemployment shock

Note: Percent of the sample noted by dashed lines at maximum household income for those in the bottom 20%, the sample median, and the min household income
for those in the top 20%. Comparable 2015 median household incomes reported for the U.S. by the Census Bureau can be found here: https://goo.gl/XkzVMR.

- 40 -
in purchasing power to avoid the same thing. 22 This result is consistent with

results from the literature summarized by Amaral (2017) -- lower income families

are more cash dependent, and thus are hurt more from an unanticipated increase in

inflation; higher income families are more likely to be borrowers, thus benefiting

from an unanticipated increase in inflation; and families in the top quartiles rely

more on earnings as an income source, thus making the potential loss of earnings

through a rise in the unemployment rate more painful. While our consideration of

a loss in purchasing power is not the same thing as a rise in inflation, the short-

term effect of a shock might be argued to have similar welfare implications.

5. Conclusions and Policy Implications

This paper presents evidence that, on average, the expected loss to family

welfare of a one-percentage point rise in the aggregate unemployment rate is

equivalent to an annualized dollar amount of $1,156. We also find a considerable

amount of heterogeneity across families, which means that aggregate averages

yield very different answers than looking more closely at population sub-groups.

For example, the expected welfare cost of a shock to unemployment is much

22
This comparison is made separately for married and single families in online
Appendix H. In spite of the dramatically different annualized dollar amount of
expected loss from a one percentage point rise in the unemployment rate, and
hence tolerance for a loss in purchasing power, the ratio of the two across the
income distribution is similar across family types.

- 41 -
higher among married (vs. single) families and increases for both in education and

income levels.

We also find that a loss in purchasing power of about 1.8%, on average for

all families, produces a welfare loss equivalent to that generated by a one

percentage point shock to unemployment and is much lower for single families

than for married families. On average singles would only be willing to trade a loss

in purchasing power of roughly one-third of a percent to avoid a one percentage

point rise in the unemployment rate, whereas married families would tolerate a

loss in purchasing power of up to three percent to avoid the same degree of

unemployment rate shock.

Additionally, we find that higher income families are willing to endure a

greater loss in purchasing power to avoid a rise in unemployment than lower

income families. This conclusion holds for both married and single families,

suggesting that, regardless of family structure and overall dollar equivalent value

of the expected loss from a rise in the unemployment rate, the willingness to

endure a loss in purchasing power to avoid higher unemployment rises

consistently across the income distribution.

The distributional consequences of monetary policy were publicly on the

minds of the U.S. Federal Open Market Committee (FOMC) in 2019. During this

year members of the FOMC participated in a series of "Fed Listens" public events

- 42 -
as part of its monetary policy framework review.23 One of the considerations of

this review includes the distributional consequences of monetary policy (see

Clarida, 2019). The analysis in this paper affirms other results in a long-standing

literature showing that monetary policy has differential impacts across the income

distribution (for example, see Coibion et al. 2017; Krusell et al. 2009). The results

here are driven by differences across households in the probability of being hit by

unemployment that might arise from contractionary monetary policy, differences

in earnings potentially lost, and by differences in the valuation of leisure by

different members of a household.

The Fed's Dual Mandate essentially requires that those who suffer the

most from the consequences of contractionary monetary policy are at least being

taken into consideration. However, a low-interest rate environment (as existed in

the U.S. in 2019) may make this more difficult. As Coibion et al. (2017) point out,

interest rates bounded by zero effectively become contractionary if economic

conditions say they should be lower. Consequently, avoiding the zero-bound, or

appealing to non-traditional monetary policy strategies, may be required in such

an environment to effectively adhere to the directive of the Dual Mandate.

23
For information about this review process, see
https://www.federalreserve.gov/monetarypolicy/review-of-monetary-policy-
strategy-tools-and-communications-fed-listens-events.htm.

- 43 -
References
Aaronson, Daniel, Sumit Agarwal, Julie L. Hotchkiss, and Taylor Kelley. 2019. ―Job
Displacement and Financial Outcomes.‖ Economics Letters 177 (C): 18–21.
Aaronson, Daniel, and Eric French. 2009. ―The Effects of Progressive Taxation on Labor Supply
When Hours and Wages Are Jointly Determined.‖ The Journal of Human Resources 44
(2): 386–408.
Alm, James, Leslie A. Whittington, and Jason Fletcher. 2002. ―Is There a ‗Singles Tax‘? The
Relative Income Tax Treatment of Single Households.‖ Public Budgeting & Finance 22
(2): 69–86. https://doi.org/10.1111/1540-5850.00074.
Amaral, Pedro. 2017. ―Monetary Policy and Inequality.‖ Cleveland Fed. January 10, 2017.
https://www.clevelandfed.org:443/newsroom and events/publications/economic
commentary/2017 economic commentaries/ec 201701 monetary policy and inequality.
Amemiya, Takeshi. 1974. ―Multivariate Regression and Simultaneous Equation Models When
the Dependent Variables Are Truncated Normal.‖ Econometrica 42 (6): 999–1012.
https://doi.org/10.2307/1914214.
Anderson, Patricia M., and Bruce D. Meyer. 1997. ―Unemployment Insurance Takeup Rates and
the After-Tax Value of Benefits.‖ The Quarterly Journal of Economics 112 (3): 913–37.
Apps, Patricia, and Ray Rees. 2009. Public Economics and the Household. Cambridge, UK:
Cambridge University Press.
Aruoba, S. Borağan, Morris A. Davis, and Randall Wright. 2016. ―Homework in Monetary
Economics: Inflation, Home Production, and the Production of Homes.‖ Review of
Economic Dynamics 21 (July): 105–24. https://doi.org/10.1016/j.red.2014.09.009.
Bahl, Roy, Richard Hawkins, Robert E. Moore, and David L. Sjoquist. 1993. ―Using
Microsimulation Models for Revenue Forecasting in Developing Countries.‖ Public
Budgeting and Financial Management 5 (1): 159–86.
Barlevy, Gadi. 2005. ―The Cost of Business Cycles and the Benefits of Stabilization.‖ Economic
Perspectives, no. Q I: 32–49.
Berentsen, Aleksander, Guido Menzio, and Randall Wright. 2011. ―Inflation and Unemployment
in the Long Run.‖ American Economic Review 101 (1): 371–98.
https://doi.org/10.1257/aer.101.1.371.
Bernanke, Ben S. 2015. ―Monetary Policy and Inequality.‖ Brookings (blog). June 1, 2015.
https://www.brookings.edu/blog/ben-bernanke/2015/06/01/monetary-policy-and-
inequality/.
Bishop, Kelly, Bradley Heim, and Kata Mihaly. 2009. ―Single Women‘s Labor Supply
Elasticities: Trends and Policy Implications.‖ Industrial and Labor Relations Review 63
(1): 146–68.
Blank, Rebecca M., and David E. Card. 1991. ―Recent Trends in Insured and Uninsured
Unemployment: Is There an Explanation?‖ The Quarterly Journal of Economics 106 (4):
1157–89. https://doi.org/10.2307/2937960.
Blau, Francine D., and Lawrence M. Kahn. 2007. ―Changes in the Labor Supply Behavior of
Married Women: 1980–2000.‖ Journal of Labor Economics 25 (3): 393–438.
https://doi.org/10.1086/513416.
Blundell, Richard. 1992. ―Labour Supply and Taxation: A Survey.‖ Fiscal Studies 13 (3): 15–40.
https://doi.org/10.1111/j.1475-5890.1992.tb00181.x.

- H1 -
Blundell, Richard, Pierre-Andre Chiappori, Thierry Magnac, and Costas Meghir. 2007.
―Collective Labour Supply: Heterogeneity and Non-Participation.‖ The Review of
Economic Studies 74 (2): 417–45.
Blundell, Richard, Alan Duncan, Julian McCRAE, and Costas Meghir. 2000. ―The Labour
Market Impact of the Working Families‘ Tax Credit.‖ Fiscal Studies 21 (1): 75–104.
https://doi.org/10.1111/j.1475-5890.2000.tb00581.x.
Blundell, Richard, and Thomas Macurdy. 1999. ―Labor Supply: A Review of Alternative
Approaches.‖ In Handbook of Labor Economics, Orley Ashenfelter and David Card,
3A:1559–1695. Amsterdam: Elsevier. https://ideas.repec.org/h/eee/labchp/3-27.html.
Blundell, Richard, Costas Meghir, Elizabeth Symons, and Ian Walker. 1988. ―Labour Supply
Specification and the Evaluation of Tax Reforms.‖ Journal of Public Economics 36 (1):
23–52. https://doi.org/10.1016/0047-2727(88)90021-7.
Browning, Martin, Pierre Chiappori, and Valérie Lechene. 2006. ―Collective and Unitary
Models: A Clarification.‖ Review of Economics of the Household 4 (1): 5–14.
Burdett, Kenneth, Mei Dong, Ling Sun, and Randall Wright. 2016. ―Marriage, Markets, and
Money: A Coasian Theory of Household Formation.‖ International Economic Review 57
(2): 337–68. https://doi.org/10.1111/iere.12160.
Carpenter, Seth B., and William M. Rodgers III. 2004. ―The Disparate Labor Market Impacts of
Monetary Policy.‖ Journal of Policy Analysis & Management 23 (4): 813–30.
https://doi.org/10.1002/pam.20048.
Chetty, Raj. 2008. ―Moral Hazard versus Liquidity and Optimal Unemployment Insurance.‖
Journal of Political Economy 116 (2): 173–234.
———. 2012. ―Bounds on Elasticities With Optimization Frictions: A Synthesis of Micro and
Macro Evidence on Labor Supply.‖ Econometrica 80 (3): 969–1018.
https://doi.org/10.3982/ECTA9043.
Clarida, Richard H. 2019. ―Speech by Vice Chair Clarida on the Federal Reserve‘s Review of Its
Monetary Policy Strategy, Tools, and Communication Practices.‖ Board of Governors of
the Federal Reserve System. April 9, 2019.
https://www.federalreserve.gov/newsevents/speech/clarida20190409a.htm.
Coibion, Olivier, Yuriy Gorodnichenko, Lorenz Kueng, and John Silvia. 2017. ―Innocent
Bystanders? Monetary Policy and Inequality.‖ Journal of Monetary Economics 88 (June):
70–89. https://doi.org/10.1016/j.jmoneco.2017.05.005.
Cullen, Julie Berry, and Jonathan Gruber. 2000. ―Does Unemployment Insurance Crowd out
Spousal Labor Supply?‖ Journal of Labor Economics 18 (3): 546–72.
https://doi.org/10.1086/209969.
Currie, Janet. 2006. ―The Take-up of Social Benefits.‖ In Poverty, the Distribution of Income,
and Public Policy, A. Auerbach, D. Card, J. Quigley (Eds.), 80–148. New York: Russell
Sage.
De Neve, Jan-Emmanuel, George Ward, Femke De Keulenaer, Bert Van Landeghem, Georgios
Kavetsos, and Michael I. Norton. 2017. ―The Asymmetric Experience of Positive and
Negative Economic Growth: Global Evidence Using Subjective Well-Being Data.‖ The
Review of Economics and Statistics, August. https://doi.org/10.1162/REST_a_00697.
Deaton, Angus. 2018. ―What Do Self-Reports of Wellbeing Say about Life-Cycle Theory and
Policy?‖ Working Paper 24369. National Bureau of Economic Research.
https://doi.org/10.3386/w24369.

- H2 -
Devereux, Paul J. 2004. ―Changes in Relative Wages and Family Labor Supply.‖ The Journal of
Human Resources 39 (3): 696–722. https://doi.org/10.2307/3558993.
DiTella, Rafael, Robert J. MacCulloch, and Andrew J. Oswald. 2001. ―Preferences over Inflation
and Unemployment: Evidence from Surveys of Happiness.‖ The American Economic
Review 91 (1): 335–41.
Dong, Mei, Ling Sun, and Randall Wright. 2015. ―Macroeconomic Policy and Household
Economics.‖ Working Paper. Economic Policy Papers. Minneapolis, MN: Federal
Reserve Bank of Minneapolis. https://www.minneapolisfed.org/research/economic-
policy-papers/macroeconomic-policy-and-household-economics.
Ebenstein, Avraham, and Kevin Stange. 2010. ―Does Inconvenience Explain Low Take-up?
Evidence from Unemployment Insurance.‖ Journal of Policy Analysis and Management
29 (1): 111–36. https://doi.org/10.1002/pam.20481.
Edwards, Kathryn Anne. 2015. ―Who Helps the Unemployed? Young Workers‘ Receipt of
Private Cash Transfers.‖ SSRN Scholarly Paper ID 2698344. Rochester, NY: Social
Science Research Network. http://papers.ssrn.com/abstract=2698344.
Eissa, Nada, Henrik Jacobsen Kleven, and Claus Thustrup Kreiner. 2008. ―Evaluation of Four
Tax Reforms in the United States: Labor Supply and Welfare Effects for Single
Mothers.‖ Journal of Public Economics 92 (3–4): 795–816.
https://doi.org/10.1016/j.jpubeco.2007.08.005.
Federal Reserve Bank of Philadelphia. n.d. ―Fourth Quarter 2016 Survey of Professional
Forecasters - Philadelphia Fed.‖ Accessed February 14, 2017.
https://www.philadelphiafed.org/research-and-data/real-time-center/survey-of-
professional-forecasters/2016/survq416.
Feenberg, Daniel, and Elisabeth Coutts. 1993. ―An Introduction to the TAXSIM Model.‖
Journal of Policy Analysis and Management 12 (1): 189–94.
https://doi.org/10.2307/3325474.
Fiorio, Carlo V. 2008. ―Analysing Tax—Benefit Reforms Using Non-Parametric Methods.‖
Fiscal Studies 29 (4): 499–522.
Flood, Sarah, Miriam King, Steven Ruggles, and Robert Warren. 2015. ―Integrated Public Use
Microdata Series, Current Population Survey: Version 4.0.‖ [Dataset]. Minneapolis:
University of Minnesota. http://doi.org/10.18128/D030.V4.0.
Fuller, David L., B. Ravikumar, and Yuzhe Zhang. 2012. ―Unemployment Insurance: Payments,
Overpayments and Unclaimed Benefits.‖ The Regional Economist. October 2012.
https://www.stlouisfed.org/~/media/Files/PDFs/publications/pub_assets/pdf/re/2012/d/Un
employment.pdf.
Gathmann, Christina, Kristiiina Huttunen, Laura Jenstrom, and Robin Stitzing. 2018. ―Job Loss
and Health Spillovers in the Family.‖ Working Paper. Finland: Aalto University.
Golberstein, Ezra, Gilbert Gonzales, and Ellen Meara. 2016. ―Economic Conditions and
Children‘s Mental Health.‖ Working Paper 22459. National Bureau of Economic
Research. http://www.nber.org/papers/w22459.
Gordon, Robert J., William Nordhaus, and William Poole. 1973. ―The Welfare Cost of Higher
Unemployment.‖ Brookings Papers on Economic Activity 1973 (1): 133–205.
https://doi.org/10.2307/2534086.
Gramlich, Edward M. 1974. ―The Distributional Effects of Higher Unemployment.‖ Brookings
Papers on Economic Activity 1974 (2): 293–336. https://doi.org/10.2307/2534189.

- H3 -
Gustman, Alan L. 1983. ―Modeling Individuals‘ Behavior: EVALUATION OF A
POLICYMAKER‘S TOOL.‖ Journal of Policy Analysis & Management 3 (2): 191–205.
Hall, Robert E. 1973. ―Wages, Income, and Hours of Work in the U.S. Labor Force.‖ In Income
Maintenance and Labor Supply, edited by Glen G. Cain and Harold W. Watts, 102–62.
Chicago, IL: University of Chicago Press.
———. 1981. ―Comment on S. Fischer ‗Relative Shocks, Relative Price Variability, and
Inflation.‘‖ Brookings Papers on Economic Activity 2: 432–34.
Heckman, James. 1974. ―Shadow Prices, Market Wages, and Labor Supply.‖ Econometrica 42
(4): 679–94. https://doi.org/10.2307/1913937.
Heim, Bradley T. 2009. ―Structural Estimation of Family Labor Supply with Taxes: Estimating a
Continuous Hours Model Using a Direct Utility Specification.‖ Journal of Human
Resources 44 (2): 350–85. https://doi.org/10.1353/jhr.2009.0002.
Herrnstein, Richard J., and Charles A. Murray. 1994. The Bell Curve: Intelligence and Class
Structure in American Life. New York: Free Press.
Hotchkiss, Julie L., Mary Mathewes Kassis, and Robert E. Moore. 1997. ―Running Hard and
Falling behind: A Welfare Analysis of Two-Earner Families.‖ Journal of Population
Economics 10 (3): 237–50.
Hotchkiss, Julie L., Robert E. Moore, and Fernando Rios-Avila. 2012. ―Assessing the Welfare
Impact of Tax Reform: A Case Study of the 2001 U.S. Tax Cut.‖ Review of Income and
Wealth 58 (2): 233–56. https://doi.org/10.1111/j.1475-4991.2012.00493.x.
Hurd, Michael. 1980. ―A Compensation Measure of the Cost of Unemployment to the
Unemployed.‖ The Quarterly Journal of Economics 95 (2): 225–43.
https://doi.org/10.2307/1885497.
Imai, Susumu, and Michael P. Keane. May2004. ―Intertemporal Labor Supply and Human
Capital Accumulation.‖ International Economic Review 45 (2): 601–41.
Kahn, Shulamit. 1997. ―Evidence of Nominal Wage Stickiness from Microdata.‖ The American
Economic Review 87 (5): 993–1008.
Keane, Michael P. 2011. ―Labor Supply and Taxes: A Survey.‖ Journal of Economic Literature
49 (4): 961–1075. https://doi.org/10.1257/jel.49.4.961.
Keane, Michael P., and Nada Wasi. 2016. ―Labour Supply: The Roles of Human Capital and The
Extensive Margin.‖ The Economic Journal 126 (592): 578–617.
https://doi.org/10.1111/ecoj.12362.
Kettemann, Adreas. 2014. ―Unemployment Insurance Takeup and the Business Cycle.‖ Working
Paper. Zurich, Germany: University of Zurich.
http://www.iza.org/conference_files/wagerigidities_2014/kettemann_a10097.pdf.
Krueger, Dirk, Kurt Mitman, and Fabrizio Perri. 2016. ―On the Distribution of the Welfare
Losses of Large Recessions.‖ Working Paper 22458. National Bureau of Economic
Research. http://www.nber.org/papers/w22458.
Krusell, Per, Toshihiko Mukoyama, Aysegul Sahin, and Jr Anthony A. Smith. 2009. ―Revisiting
the Welfare Effects of Eliminating Business Cycles.‖ Review of Economic Dynamics 12
(3): 393–402.
Krusell, Per, and Anthony A Smith. 1999. ―On the Welfare Effects of Eliminating Business
Cycles.‖ Review of Economic Dynamics 2 (1): 245–72.
https://doi.org/10.1006/redy.1998.0043.

- H4 -
Kumar, Anil. 2009. ―Nonparametric Estimation of the Impact of Taxes on Female Labor
Supply.‖ Journal of Applied Econometrics 27 (3): 415–39.
https://doi.org/10.1002/jae.1205.
Lagos, Ricardo, Guillaume Rocheteau, and Randall Wright. Forthcoming. ―Liquidity: A New
Monetarist Perspective.‖ Journal of Economic Literature.
http://econweb.ucsd.edu/~rstarr/281webpage/Lagos_Rochetau_Wright.pdf.
Lam, David. 1988. ―Marriage Markets and Assortative Mating with Household Public Goods:
Theoretical Results and Empirical Implications.‖ The Journal of Human Resources 23
(4): 462–87. https://doi.org/10.2307/145809.
Lucas, Robert E. 1991. Models of Business Cycles. Oxford, OX, UK; Cambridge, Mass., USA:
Wiley-Blackwell.
Mathers, C.D., and D.J. Schofield. 1998. ―The Health Consequences of Unemployment: The
Evidence.‖ The Medical Journal of Australia 168 (4): 178–82.
McClelland, Robert, and Shannon Mok. 2012. ―A Review of Recent Research on Labor Supply
Elasticities.‖ Working Paper 2012–12. Washington, D.C.: Congressional Budget Office.
McElroy, Marjorie B. 1990. ―The Empirical Content of Nash-Bargained Household Behavior.‖
The Journal of Human Resources 25 (4): 559–83. https://doi.org/10.2307/145667.
Michaelides, Marios, and Peter R. Mueser. 2012. ―Recent Trends in the Characteristics of
Unemployment Insurance Recipients.‖ Monthly Labor Review 135 (7): 28–47.
Moreau, Nicolas, and Olivier Bargain. 2005. ―Is the Collective Model of Labor Supply Useful
for Tax Policy Analysis? A Simulation Exercise.‖ SSRN Scholarly Paper ID 396040.
Rochester, NY: Social Science Research Network.
http://papers.ssrn.com/abstract=396040.
Mroz, Thomas A. 1987. ―The Sensitivity of an Empirical Model of Married Women‘s Hours of
Work to Economic and Statistical Assumptions.‖ Econometrica 55 (4): 765–99.
https://doi.org/10.2307/1911029.
Mukoyama, Toshihiko, and Ayşegül Şahin. 2006. ―Costs of Business Cycles for Unskilled
Workers.‖ Journal of Monetary Economics 53 (8): 2179–93.
https://doi.org/10.1016/j.jmoneco.2005.08.014.
Muthen, Bengt. 1990. ―Moments of the Censored and Truncated Bivariate Normal Distribution.‖
British Journal of Mathematical and Statistical Psychology 43 (1): 131–43.
https://doi.org/10.1111/j.2044-8317.1990.tb00930.x.
Nakajima, Makoto. 2015. ―Te Redistributive Consequences of Monetary Policy.‖ Business
Review. Federal Reserve Bank of Philadelphia. https://www.philadelphiafed.org/-
/media/research-and-data/publications/business-
review/2015/q2/brQ215_the_redistributive_consequences_of_monetary_policy.pdf.
Okun, Arthur M. 1962. ―Potential GNP: Its Measurement and Significance.‖ Proceedings of the
Business and Economics Statistics Section of the Am Statistical Assoc, 98–104.
Pencavel, John. 2002. ―A Cohort Analysis of the Association between Work Hours and Wages
among Men.‖ The Journal of Human Resources 37 (2): 251–74.
https://doi.org/10.2307/3069647.
Proto, Eugenio, and Aldo Rustichini. 2013. ―A Reassessment of the Relationship between GDP
and Life Satisfaction.‖ PLOS ONE 8 (11): e79358.
https://doi.org/10.1371/journal.pone.0079358.

- H5 -
Ransom, Michael R. 1987. ―An Empirical Model of Discrete and Continuous Choice in Family
Labor Supply.‖ The Review of Economics and Statistics 69 (3): 465–72.
https://doi.org/10.2307/1925534.
Soest, Arthur van. 1995. ―Structural Models of Family Labor Supply: A Discrete Choice
Approach.‖ The Journal of Human Resources 30 (1): 63–88.
https://doi.org/10.2307/146191.
Stevenson, Betsey, and Justin Wolfers. 2013. ―Subjective Well-Being and Income: Is There Any
Evidence of Satiation?‖ American Economic Review Papers and Proceedings 103 (3):
598–604. https://doi.org/10.1257/aer.103.3.598.
Sullivan, Daniel, and Till von Wachter. 2009. ―Job Displacement and Mortality: An Analysis
Using Administrative Data.‖ The Quarterly Journal of Economics 124 (3): 1265–1306.
https://doi.org/10.1162/qjec.2009.124.3.1265.
Triest, Robert K. 1990. ―The Effect of Income Taxation on Labor Supply in the United States.‖
The Journal of Human Resources 25 (3): 491–516. https://doi.org/10.2307/145991.

- H6 -