The effect of minimum wages on employment in the presence of

∗
productivity fluctuations
Asahi Sato†

February 27, 2025

arXiv:2502.18261v2 [econ.GN] 26 Feb 2025

Abstract

Traditionally, the impact of minimum wages on employment has been studied, and it is
generally believed to have a negative effect. Yet, some recent studies have shown that the
impact of minimum wages on employment can sometimes be positive. In addition, certain
recent proposals set a higher minimum wage than the wage earned by some high-productivity
workers. However, the impact of minimum wages on employment has been primarily stud-
ied on low-skilled workers, whereas there is limited research on high-skilled workers. To
address this gap and examine the effects of minimum wages on high-productivity workers’
employment, I construct a macroeconomic model incorporating productivity fluctuations,
incomplete markets, directed search, and on-the-job search and compare the steady-state
distributions between the baseline model and the model with a minimum wage. As a result,
binding minimum wages increase the unemployment rate of both low and high-productivity
workers.

JEL Classification Numbers: E21, E24, J60.

Keywords: Minimum wage; On-the-job search; Directed search.

1 Introduction

In recent years, the issue of raising the minimum wage has been a topic of extensive dis-
cussion. Introducing or raising minimum wages has various economic effects. In particular, the
effects on employment have long been a subject of traditional research and many empirical stud-
ies show that minimum wages have adverse impacts on employment (Neumark and Nizalova,
2007; Eckstein, Ge and Petrongolo, 2011; Sabia, Burkhauser and Hansen, 2012). Yet, some re-
cent studies are critical of the adverse effects of minimum wages on employment (Cengiz et al.,
2019; Dustmann et al., 2022; Engbom and Moser, 2022). Thus, the effects of minimum wages
on employment remain a subject of ongoing debate. Though the effects of minimum wages are
∗
I am grateful to Gaston Chaumont for answering a wide range of questions and to Shinichi Nishiyama and
Naoto Jinji for providing valuable comments and suggestions.
†
Fourth-year student, Faculty of Economics, Kyoto University (email: [email protected]).

1
unclear, the increase in minimum wages has been steadily proceeding. For example, as pointed
out in Hurst et al. (2022), in the United States, a proposal to raise the minimum wage from
$7.25 to $15 is being considered and about 40% of non-college-educated workers and 10% of
college-educated workers earn a wage lower than $15. This implies that minimum wages have
direct effects on not only low-productivity workers but also high-productivity workers. How-
ever, most research examining the employment effects of minimum wages focuses on teenagers,
low-wage industries, and workers with low education levels (Manning, 2021). This implies that
only a few studies focus on high-productivity workers.
Motivated by this, I examine the effects of minimum wages on the employment of high-
productivity workers by using a macroeconomic model that incorporates productivity fluctua-
tions and incomplete markets a la Aiyagari (1994), directed search, and on-the-job search. In
constructing the model for this paper, I referred to several papers. Hasumi and Takano (2024)
incorporate Aiyagari (1994) type heterogeneous households and assume that wages are deter-
mined by bargaining. The main difference between their research and mine is the labor market
friction incorporated. I choose directed search as I consider it to be more realistic in terms of
households’ knowing wages before applications. Zhao and Sun (2021) add minimum wages to
the model used in Tsuyuhara (2016), which incorporates directed search and on-the-job search.
This research is very similar to mine, but in this research, households are assumed not to save.
I think that this is unrealistic so the model I use allows for household savings.
This paper makes two contributions. One is incorporating productivity fluctuations a la
Aiyagari (1994) into a model that adopts the block recursive equilibrium. As far as I know,
no other paper combines these two. The other is demonstrating that minimum wages raise the
unemployment rate of both low and high-productivity workers.
The next section describes the model proposed in this paper. Section 3 presents a quantitative
analysis. The final section concludes.

2 Model

2.1 Environment of the model

I develop the baseline model based on the model used in Chaumont and Shi (2022) and
Aiyagari (1994). However, for simplicity, I omit the assumption that unemployment insurance
is calculated based on wages from the former position, which is adopted in Chaumont and Shi
(2022).
Time is discrete and infinite. There is a unit measure of risk averse workers and risk neutral
firms whose measure is determined by competitive entry. Each worker has a period utility

2
function u(·) defined over consumption c which satisfies u′ (c) ∈ (0, ∞) and u′′ (c) ∈ (−∞, 0) for
∀c ∈ R+ , with u′ (0) = ∞. Workers and firms have the same discount factor β ∈ (0, 1). Workers
can accumulate non-contingent assets denoted as a ≥ a, where a is the lower bound on asset
holdings. When a is negative, it is called a borrowing limit. The net rate of return on assets r
is taken as given in this model. Ex ante identical workers become endogenously heterogeneous
ex post in terms of their labor market status, their productivity, current earnings, and wealth.
A worker’s labor market status is represented as ε ∈ {e, u}, where e means being employed,
and u means being unemployed. A worker’s productivity is η ∈ ll , lh , where ll denotes low-


productivity and lh denotes high-productivity. Denote a worker’s earnings as ω. For an employed

worker, ω is equal to the worker’s wage w. For an unemployed worker, ω is equal to 0.
Each entering firm is endowed with a job and hires a worker. The production technology
has constant returns to scale in jobs, and a firm treats the jobs independently. I refer to a job
as a firm. Depending on the worker’s productivity η, jobs yield a stream of output ηy, and the
production cost is normalized to 0. Firms commit to offers and cannot respond to the employee’s
outside offers so if an employed worker receives a better offer, the worker quits the current job.
In addition to this endogenous separation, exogenous separation destroys an existing match with
probability δ ∈ (0, 1). Exogenous separation is independent across matches and time.
In each period, an unemployed worker searches with a probability of 1, while an employed
worker with a probability of λe ∈ (0, 1). The labor market consists of a continuum of submarkets,
which are indexed by the applicant’s productivity η, the wage offer w, and the applicant’s
wealth a. Each submarket (η, w, a) has a tightness θ(η, w, a), which is the ratio of vacancies
to applicants. Taking the tightness function as given, firms and workers choose the submarket
to enter. Workers who search in submarket (η, w, a) must be endowed with productivity η and
have the wealth level a. It may seem unnatural that applicants’ wealth is one of the factors
determining the ratio of vacancies to applicants. However, incorporating applicants’ wealth as
an argument of market tightness simplifies the calculation of equilibrium.
In any submarket with tightness θ, the matching probability is p(θ) for an applicant and q(θ)
for a vacancy, where p(θ) = θq(θ). Furthermore, the matching probabilities satisfy the following
assumptions: p(θ), q(θ) ∈ [0, 1], p′ (θ) ∈ (0, ∞), q ′ (θ) ∈ (−∞, 0) and p′′ (θ) < 0 for ∀ θ ≥ 0.
Each period is divided into five stages: (i) production, (ii) consumption and savings, (iii)
search and matching, (iv) exogenous separation, and (v) productivity fluctuations. In stage
(i), all existing matches produce, and workers obtain earnings. In stage (ii), workers make
decisions on consumption and savings. In stage (iii), the opportunity to search is realized with
the probability λe for each employed worker and with probability 1 for each unemployed worker.
In stage (iv), a separation shock destroys a fraction δ ∈ (0, 1) of old matches and throws the

3
workers into unemployment. However, I assume that a newly formed match is not subject to
exogenous separation until the next period. In stage (v), each worker’s productivity fluctuates
according to the 2-state Markov chain which is explained later in section 3.1.
A worker’s individual state consists of the labor market status ε, current productivity η,
current earnings ω, and wealth a. I express it as s ≡ (ε, η, ω, a). The labor market status is an
element of the set E ≡ {e, u}. The set L ≡ ll , lh contains the worker’s productivity. Earnings


are drawn from the set W ≡ [w, w̄] for an employed worker and in the set {0} for an unemployed
worker. Denote W̄ ≡ W ∪ {0}.Wealth belongs to the set A ≡ [a, ā]. Define the Borel sets as
E corresponding to E, L corresponding to L, W corresponding to W̄ , and A corresponding to
A. The individual state space is defined as S ≡ E × L × W̄ × A with its associated Borel sets
S = E × L × W × A.
The aggregate state of the economy is represented by ψ : S → [0, 1], which describes a distri-
bution of workers over individual states. The space of distribution functions on the measurable
space (S, S) is denoted by Ψ(S, S). The law of motion for the state of the economy is given by
the operator T : Ψ(S, S) → Ψ(S, S).

2.2 Decisions over consumption and savings

Consider a worker’s state is (ε, η, ω, a), then, the worker’s value function is defined as follows:

Vε (η, ω, a) = max [u(c) + βRε (η, ω, â)]

(c,â)
(1)
â
s.t. c + = ymin + ω + a and â ≥ a,
1+r

where â is the assets at the beginning of the next period1 and Rε (η, ω, â) is also a value function
about the return of search defined later. The ymin is a small amount of home production to
prevent c ≤ 0. Denote the policy function as cε (η, ω, a) for optimal consumption and âε (η, ω, a)
for optimal savings.

2.3 Unemployed workers’ search

Consider a low-productivity unemployed worker. An unemployed worker chooses a submar-

ket (ll , ŵ, â) to search in and gets a job with probability p(θ(ll , ŵ, â)). Then, the worker starts
the next period with value Ve (ll , ŵ, â) or Ve (lh , ŵ, â). If the worker fails to match, the worker
starts the next period with the value Vu (ll , 0, â) or Vu (lh , 0, â). Let Ru (η, 0, â) denote the value
generated by the optimal search. Except for the probabilities of being hired, the same can be
1
The hat is used to distinguish between current and next period.

4
said for high-productivity workers. Then,
n
Ru (ll , 0, â) = max p(θ(ll , ŵ, â))(Tll Ve (ll , ŵ, â) + Tlh Ve (lh , ŵ, â))
ŵ
o (2)
l l h
+(1 − p(θ(l , ŵ, â)))(Tll Vu (l , 0, â) + Tlh Vu (l , 0, â))

and
n
Ru (lh , 0, â) = max p(θ(lh , ŵ, â))(Thl Ve (ll , ŵ, â) + Thh Ve (lh , ŵ, â))
ŵ
o (3)
+(1 − p(θ(lh , ŵ, â)))(Thl Vu (ll , 0, â) + Thh Vu (lh , 0, â)) ,

where Tll , Tlh , Thl , and Thh are the productivity transition probabilities. The subscript ll denotes
transition from low to low productivity and lh means from low to high productivity. The same
goes for hl and hh as well. Denote the policy function as w̃u (η, 0, â) for optimal search target.
Because âu is a function of a, w̃u (η, 0, â) is also a function of a so I express the optimal search
policy as ŵu (η, 0, a) ≡ w̃u (η, 0, âu (η, 0, a)). Similarly, I express the tightness of the submarket
targeted by the unemployed worker as θ̂u (η, 0, a) ≡ θ(η, ŵu (η, 0, a), âu (η, 0, a)).

2.4 Employed workers’ search

Consider a low-productivity employed worker with wage w and wealth â. The worker can
search only when he receives an opportunity to search, which occurs with probability λe and
then, the worker chooses a submarket (ll , ŵ, â) to search. Then, the worker starts the next
period with the value Ve (ll , ŵ, â) or Ve (lh , ŵ, â). Even if search and matching fails, the worker
does not immediately lose the worker’s current job. Exogeneous separation will affect the worker
with probability δ. If the worker escapes exogeneous separation, the continuation value will be
Ve (ll , w, â) or Ve (lh , w, â). If not, the continuation value will be Vu (ll , 0, â) or Vu (lh , 0, â). Except
for the probabilities of being hired, the same can be said for high-productivity workers. Thus,
n
Re (ll , w, â) = max λe p(θ(ll , ŵ, â))(Tll Ve (ll , ŵ, â) + Tlh Ve (lh , ŵ, â))
ŵ

+ (1 − λe p(θ(ll , ŵ, â)))Tll [(1 − δ)Ve (ll , w, â) + δVu (ll , 0, â)] (4)
o
+(1 − λe p(θ(ll , ŵ, â)))Tlh [(1 − δ)Ve (lh , w, â) + δVu (lh , 0, â)]

and
n
Re (lh , w, â) = max λe p(θ(lh , ŵ, â))(Thl Ve (ll , ŵ, â) + Thh Ve (lh , ŵ, â))
ŵ

+ (1 − λe p(θ(lh , ŵ, â)))Thl [(1 − δ)Ve (ll , w, â) + δVu (ll , 0, â)] (5)
o
+(1 − λe p(θ(lh , ŵ, â)))Thh [(1 − δ)Ve (lh , w, â) + δVu (lh , 0, â)] .

5
Denote the policy function as w̃e (η, w, â) for optimal search target. I express the optimal search
target as a function of wealth at the beginning of the period a: ŵe (η, w, a) ≡ w̃e (η, w, âe (η, w, a)).
Also, denote the tightness of targeted submarket as θ̂e (η, w, a) ≡ θ(η, ŵe (η, w, a), âe (η, w, a)).

2.5 Firms and market tightness

Consider a job with wage w filled by a low-productivity worker with wealth a. The cur-
rent profits of the job are ll y − w. The worker will quit for another job with probability
λe θ̂e (ll , w, a) and also, may face exogeneous separation with probability δ. If separation oc-
curs, the future value of the job is 0. If no separation occurs, the future value of the job
is β Tll J(ll , w, âe (ll , w, a)) + Tlh J(lh , w, âe (ll , w, a)) . Except for the output and the matching


probabilities, the same can be said for high-productivity workers. The value of filled jobs is:

n o
J(ll , w, a) = ll y−w+(1−δ)[1−λe p(θ̂e (ll , w, a))]β Tll J(ll , w, âe (ll , w, a)) + Tlh J(lh , w, âe (ll , w, a))

(6)

and

n o
J(lh , w, a) = lh y−w+(1−δ)[1−λe p(θ̂e (lh , w, a))]β Thl J(ll , w, âe (lh , w, a)) + Thh J(lh , w, âe (lh , w, a)) .

(7)

Competitive entry of vacancies determines each submarket tightness. Consider submarket

(ll , ŵ, â). If the firm is matched, the expected present value is βq(θ(ll , ŵ, â))Tll J(ll , ŵ, â) +
βq(θ(ll , ŵ, â))Tlh J(lh , ŵ, â). The flow cost of a vacancy is k > 0. Thus, for all (ll , ŵ, â), if there
is potential surplus, that is, if β Tll J ll , ŵ, â + Tlh J lh , ŵ, â



≥ k, free entry of vacancies
matches the values of posting cost k and the expected present value of a vacancy. Thus, the
following equation holds:

n o
l l h
0 = −k + βq(θ(l , ŵ, â)) Tll J(l , ŵ, â) + Tlh J(l , ŵ, â) . (8)

If there is no potential surplus, that is, if k > β Tll J ll , ŵ, â + Tlh J lh , ŵ, â , no vacancy


enters the market. Then, θ ll , ŵ, â = 0. From these two, the following equation can be derived:

  
q −1 k
if β Tll J ll , ŵ, â + Tlh J lh , ŵ, â ≥ k

 


β {Tll J (ll ,ŵ,â)+Tlh J (lh ,ŵ,â)}
 
θ ll , ŵ, â = (9)

0

otherwise.

6
Same goes for θ lh , ŵ, â . Thus,

  
−1 k
if β Thl J ll , ŵ, â + Thh J lh , ŵ, â ≥ k
 


q

β {Thl J (ll ,ŵ,â)+Thh J (lh ,ŵ,â)}
 
θ lh , ŵ, â =

0

otherwise.
(10)

2.6 Equilibrium

The aggregate state of the economy is characterized by the distribution of workers across
individual states. Given the distribution of workers at the beginning of a period, ψ, individuals’
optimal choices, matching outcomes, and productivity fluctuations determine the distribution
of workers at the start of the next period, ψ̂.

Definition 1. An equilibrium consists of the value function of workers Vε (for ε = e, u), the firm
value function J, policy functions (cε , âε ) and (ŵu , ŵe ), and the transition of the aggregate state
T , such that: (i) The value functions of workers Vε : L×W̄ ×A → R satisfies (1), optimal decisions
of consumption and savings yield the policy functions (cε , âε ), and optimal search decisions yield
the policy functions (ŵu , ŵe ); (ii) the firm value function J : L×W ×A → R satisfies (6) and (7);
(iii) the tightness function θ satisfies (9) and (10) for all (ll , ŵ, â), (lh , ŵ, â) ∈ L × W × A; and
(iv) the aggregate state transition T : Ψ(S, S) → Ψ(S, S) is consistent with the policy functions.

This equilibrium is a block recursive equilibrium (BRE), which is defined and analyzed by
Shi (2009); Menzio and Shi (2010, 2011). In this equilibrium, the distribution of workers does
not influence value functions, policy functions, and the market tightness function (see section 3
in Chaumont and Shi (2022)).

3 Quantitative analysis

I choose parameters to reproduce the unemployment rate and labor market transition rates
in Japan for 2023.

3.1 Calibration

The utility function have the following forms:

c1−σ
u(c) = . (11)
1−σ

7
 Table 1: Main parameter values.

Parameters Descriptions Values Comments / Targets

β discount factor 0.996 when raised to the power of three, this
is ≒ 0.987 used in Menzio and Shi
(2010)
r the net rate of return on assets 0.327% annual interest rate = 4.0%
a borrowing limit -5.0 no specific reason
δ exogeneous separation rate 0.633% calculated based on the data from
Labour Force Survey
σ the risk aversion parameter 2.0 often used in macro calibration
γ the matching technology parameter 0.5 following Menzio and Shi (2010)
λe the probability of having the opportu- 0.0346 job-to-job transition rate = 0.41%
nity to search
k the cost of posting a job 0.77 unemployment rate = 2.6%

Following Menzio and Shi (2010), the matching probabilities have the following forms:

− γ1 p(θ)
p(θ) = θ(1 + θγ ) , q(θ) = . (12)
θ

I calibrate the model to the monthly frequency. Normalize y = 1. Table1 lists the parameters
and calibration targets. The monthly interest rate is r = 0.327%, or 4.0% annually. I set β
to 0.996. When raised to the power of three, this is ≒ 0.987, which is used in Menzio and Shi
(2010)2 . I set a to -5.0. The risk aversion parameter is set to σ = 2, which is commonly used in
the literature.
I set the exogenous separation rate δ = 0.00633 ≒ 430000/(430000+67470000). The value
430000 is the number of involuntary job separation and the value 67470000 is the number
of employees. Both values are from 2023 Yearly Average Results of Labour Force Survey
(Statistics Bureau of Japan, 2024b). For an employed worker, the probability of having the
opportunity to search λe is chosen to yield a monthly job-to-job transition rate of 0.41% ≒
(3280000/12)/67470000. The value 3280000 is the number of job changers during 2023 taken
from 2023 Yearly Average Results (Detailed Tabulation) of Labour Force Survey (Statistics
Bureau of Japan, 2024a). Following Menzio and Shi (2010), the matching technology param-
eter γ is set to be 0.5. Following Chaumont and Shi (2022), unearned income ymin is set to
be 0.05. The vacancy cost k is set to be 0.77 to match an unemployment rate of 2.6% ≒
1780000/(1780000+67470000). The value 1780000 is the number of unemployed person taken
from 2023 Yearly Average Results of Labour Force Survey (Statistics Bureau of Japan, 2024b).

2
In the paper, the period is set to be one quarter.

8
 I assume that labor productivity follow an AR(1) process as follows:

log(lt ) = ρ log(lt−1 ) + ϵt , ϵt ∼ N (0, σ22 ). (13)

I discretize it with ρ = 0.99 and σ2 = 0.05 into a 2-state Markov chain by using the method of
Rouwenhorst (1995). As a result, ll is set to be 0.7016 and lh is set to be 1.4254. Also, Tll is set
to be 0.995, Tlh is set to be 0.005, Thl is set to be 0.005, and Thh is set to be 0.995.

3.2 Solution method and stationary distribution

I define the grids for the spaces of wealth and wages and set the number of grid points to
141 and 20, respectively. Wealth is set to be in [a, ā] = [−5.0, 700] and wages are set to be in
[w, w̄] = [0, 1.4254]. To solve the model, I use the following algorithm:

1. Guess initial value functions of workers and firms.

2. Given the value functions of firms, solve for the market tightness using (9) and (10).

3. Given the tightness function, solve the worker’s optimization problem and compute optimal
savings, consumption and job search decisions. In this step, I use Value Function Iteration,
that is, this step is iterated until resulting value functions converge.

4. Given worker’s policy functions computed in the previous step, calculate separation rates
of employed workers and update the value function of the firm in each submarket. Until
convergence, I iterate this step.

5. Check the difference between initial value functions of firm and previously computed ones.
If the difference is sufficiently small, end this algorithm. If not, using calculated value
functions instead of initial guess, repeat steps 2 to 4 above.

After solving the model, I derive the transition probability matrix for household state vari-
ables using policy functions, and also, I obtain stationary distribution by use of this transition
probability matrix.

3.3 Results in the baseline model

Table 2 shows the main resulting indicator values in the baseline model. The unemployment
rate for low-productivity workers is 1.396%, which is higher than that for high-productivity
workers. The job-to-job transition rate in the baseline model is 0.406%.
Figure 1 depicts a worker’s optimal search decisions without minimum wages as a function
of current wealth. In the legend, I use w1, w2, . . . , w20, which represent the 1st grid point,

9
 Table 2: Indicator values in the baseline model.

Indicator Values
Unemployment rate of low-productivity workers 1.396%
Unemployment rate of high-productivity workers 1.178%
Job-to-job transition rate 0.406%

2nd grid point, . . . , 20th grid point, respectively. As shown in this figure, the more wealth
workers have, the higher wage workers target. The submarket tightness θ is decreasing in wage
so that the matching probability is also decreasing in wage. Because of this, workers with
enough wealth tend to run the risk of losing a job by applying for a higher-paying job. When
comparing high-productivity workers and low-productivity workers, high-productivity workers
are predisposed to apply for a job with a higher wage. When other arguments (ŵ and â) are the
same, θ(lh , ŵ, â) ≥ θ(ll , ŵ, â) is always valid because the value of a job with a high-productivity
worker is higher than that with a low-productivity worker. That is why, a high-productivity
worker apply for a higher-paying job. From panels (b) and (c), we can see that as the wage of
the current job increases, workers target higher wages. This is because there is no reason for an
employed worker to apply for a job with lower wages than those of the worker’s current job. In
panel (b), it seems that the lines indexed by w12, w13, . . . , w19 are not displayed. Likewise,
in panel (c), the lines indexed by w18 and w19 do not seem to be displayed. They are hidden
under the line “w20”. This phenomenon does not stem from a mistake in my code and is not an
issue at all. The vacancy post cost k is so high relative to output that as shown in Figure 6, the
market tightnes falls to zero quickly. Because of this, the market tightness of a job for which a
worker with highly-paid job want to apply is 0. Then, for employed workers with well-paying
job, it does not matter which wages they target.
Figure 2 depicts the difference between next period’s wealth and current wealth without
minimum wages. From panel (a), it can be understood that high-productivity unemployed
workers tend to save less wealth. They will earn more relative to low-productivity workers
with high probabilities so that they consume more. Panels (b) and (c) show that there are
no significant differences in the saving policy depending on the productivity of an employed
worker. Moreover, from this figure, we can see that saving is a decreasing function of wealth at
the beginning of the period. If the amount of wealth at the beginning of the period is small,
it is very important for workers to save more because they have more chances of getting hired.
However, if a worker has a sufficient amount of wealth at the beginning of the period, it is
optimal for the worker to save less and gain instant utility at the cost of wealth at the beginning

10
of the next period.
From Figure 3, we can see that the more wealth workers have, the more they consume. We
can see that from panel (a), high-productivity unemployed workers consume more. Panels (b)
and (c) show that as the wage of the current job increases, workers consume more. It is not
surprising that consumption policy is an increasing function of both wealth and productivity.
Figure 4 depicts the stationary distribution without minimum wages. Panel (a) shows that
in the equilibrium, earnings tend to be concentrated at the higher end. Ten bars3 appear in
Panel (a). The third bar from the left represents the proportion of employed workers with ”w10”
wage (w̄ × (10 − 1)/(20 − 1) ≒ 0.6752). As stated in the section 3.4, in testing the effects of
minimum wages, I set w to 0.1 and 0.7. The former expresses non-binding minimum wages,
while the latter expresses binding minimum wages. Panel (b) shows that in the equilibrium, the
distribution of wealth is characterized by a long tail on the right side.
In Figure 5, the firm value of a filled job is expressed as a function of wealth or wage. Because
there are so many grid points in the space of wealth, in panels (c) and (d), I pick up and display
lines indexed by 1st, 12th, and 141st grid point. When comparing panels (a) and (b) or (c)
and (d), it can be seen that the value of a firm with a high-productivity worker is higher. We
can see that as wages increase, the value of the job decreases. Furthermore, the job value is an
increasing function of wealth.
Figure 6 depicts market tightness as a function of wealth or wage. Because there are so
many grid points in the space of wealth, in panels (c) and (d), I pick up and display lines
indexed by 1st, 12th, and 141st grid point. When comparing panels (a) and (b) or (c) and (d),
it can be understood that the market tightness for a high-productivity worker is higher. We
can see that as wages values increase, the market tightness decreases. Furthermore, the market
tightness is an increasing function of wealth. Note that because the market tightness θ is the
ratio of vacancies to applicants, the higher the market tightness is, the more advantageous for
workers. Why is the value of job and market tightness an increasing function of wealth? As
shown in Figure 1, more wealth causes a worker to apply for a job with higher wages. This
means that it is difficult for employed workers with significant wealth to change jobs so that the
job conditioned by much wealth is less likely to expire. Then, the value of job is an increasing
function of wealth. Because the market tightness rises as the job value increases, the market
tightnes is also an increasing function of wealth.

3
Note that although it is barely visible, there is the second bar from the left to the right of the number 0.5.

11
 (a) Search Policy for Unemployed Workers

(b) Search Policy for Low-Productivity Employed Workers

(c) Search Policy for High-Productivity Employed Workers

Figure 1: Optimal search policy.

12
 (a) Saving Policy for Unemployed Workers

(b) Saving Policy for Low-Productivity Employed Workers

(c) Saving Policy for High-Productivity Employed Workers

Figure 2: Optimal saving policy.

13
 (a) Consumption Policy for Unemployed Workers

(b) Consumption Policy for Low-Productivity Employed Workers

(c) Consumption Policy for High-Productivity Employed Workers

Figure 3: Optimal consumption policy.

14
(a) Stationary Distribution of Earnings

(b) Stationary Distribution of Wealth

Figure 4: Stationary distribution.

15
(a) Firms Value with a Low-Productivity (b) Firms Value with a High-Productivity
Worker and Wealth Worker and Wealth

(c) Firms Value with a Low-Productivity (d) Firms Value with a High-Productivity
Worker and Wage Worker and Wage

Figure 5: Firms value.

(a) Market Tightness for a Low-Productivity (b) Market Tightness for a High-Productivity
Worker and Wealth Worker and Wealth

(c) Market Tightness for a Low-Productivity (d) Market Tightness for a High-Productivity
Worker and Wage Worker and Wage

Figure 6: Market tightness.

16
 Table 3: Indicator values in the baseline and with minimum wages.

Indicator Values (baseline → with the non-binding → with the binding)

Unemployment rate of the low 1.396% → 1.396% → 28.521%
Unemployment rate of the high 1.178% → 1.178% → 1.600%
Job-to-job transition rate 0.406% → 0.406% → 0.230%

3.4 Minimum wages

I solve three types of models, the baseline model, the model with the non-binding minimum
wage, and the model with the binding minimum wage. As mentioned in section 3.3, in the model
with the non-binding minimum wage, I express it by setting w to 0.1. Similarly, in the model
with the binding minimum wage, I set w to 0.7. In the remainder of this section, I compare the
results of the baseline model to those of model with non-binding or binding minimum wage.
Table 3 shows main resulting indicator values in the baseline model, the model with non-
binding minimum wages, and the model with binding minimum wages. We can see that non-
binding minimum wages have few effects on the economy. On the other hand, introducing binding
minimum wages increases the unemployment rate of both low and high-productivity workers and
decreases the job-to-job transition rate. The magnitude of increase in the unemployment rate
is higher among low-productivity workers.
Figure 7 depicts the stationary distribution of the baseline model and that of the model with
non-binding minimum wages. Blue color bars and lines represent the stationary distribution in
the baseline model. Red color represents that in the model with the minimum wage. However,
in Figure 7, most of bars and lines are overlapped so that it seems that mixed color bars and
lines are displayed. Incorporating non-binding minimum wages has few effects on the stationary
distribution.
Figure 8 depicts the stationary distribution of baseline model and that of model with binding
minimum wages. From Panel (a), we can see that the binding minimum wage increases the
unemployment rate and the number of employed workers with higher-paying jobs. Panel (b)
shows that the wealth distribution shifts upward with the introduction of the binding minimum
wage.

17
(a) Stationary Distribution of Earnings of the Baseline and with Non-Binding Minimum Wages

(b) Stationary Distribution of Wealth of the Baseline and with Non-Binding Minimum Wages

Figure 7: Stationary distribution of the baseline and with non-binding minimum wages.

18
(a) Stationary Distribution of Earnings of the Baseline and with Binding Minimum Wages

(b) Stationary Distribution of Wealth of the Baseline and with Binding Minimum Wages

Figure 8: Stationary distribution of the baseline and with binding minimum wages.

19
3.5 Discussion

As traditionally argued, introducing binding minimum wages increases the unemployment

rate. According to the results of the policy experiment, low-productivity workers are strongly af-
fected by the minimum wage. This is because the binding minimum wage I incorporate eliminates
jobs targeted only by low-productivity workers. Even though jobs targeted by high-productivity
workers are not eliminated, the unemployment rate of high-productivity workers also increases
after incorporating the minimum wage. Why are high-productivity workers also affected by
the minimum wage? Since workers’ productivity fluctuates, then, some low-productivity unem-
ployed workers affected by the minimum wage become high productivity unemployed workers.
This means that it is low-productivity workers who transmit the effect of the minimum wage on
the employment of high-productivity workers.
Additionally, by introducing the binding minimum wage, workers accumulate more wealth.
Due to the increased likelihood of unemployment, people begin saving more to smooth their
consumption over time. See Figure 9, which displays the value obtained by subtracting saving
policy in the model with the binding minimum wage from that in the baseline model. We can
see that all values are less than or equal to zero meaning that workers tend to accumulate more
wealth when the binding minimum wage is present.

20
 (a) Differences in Saving Policy for the Unemployed between Both Models

(b) Differences in Saving Policy for the Low-Productivity Employed between Both Models

(c) Differences in Saving Policy for the High-Productivity Employed between Both Models

Figure 9: Saving policy in the baseline minus that in the model with binding minimum wages.

21
4 Conclusion

I examine the effects of minimum wages on employment by use of a macroeconomic model

incorporating productivity fluctuations and incomplete markets a la Aiyagari (1994), directed
search, and on-the-job search.
I find that as traditionally argued, the binding minimum wage increases the unemployment
rate. However, not only low-productivity workers but also high-productivity workers experience
an increase in the unemployment rate. In addition, binding minimum wages drive households
to accumulate more wealth.
There are two possible extensions to this model. First, we can discretize the AR(1) process
into a Markov chain with more states. Second, we can reincorporate the omitted assumption,
that is, we can incorporate the assumption that unemployment insurance is calculated based on
the wage from the former position. I think that this assumption makes the stationary distribution
of earnings more realistic.

References

Aiyagari, S Rao. 1994. “Uninsured idiosyncratic risk and aggregate saving.” The Quarterly
Journal of Economics, 109(3): 659–684.

Cengiz, Doruk, Arindrajit Dube, Attila Lindner, and Ben Zipperer. 2019. “The effect
of minimum wages on low-wage jobs.” The Quarterly Journal of Economics, 134(3): 1405–
1454.

Chaumont, Gaston, and Shouyong Shi. 2022. “Wealth accumulation, on-the-job search
and inequality.” Journal of Monetary Economics, 128: 51–71.

Dustmann, Christian, Attila Lindner, Uta Schönberg, Matthias Umkehrer, and

Philipp Vom Berge. 2022. “Reallocation effects of the minimum wage.” The Quarterly
Journal of Economics, 137(1): 267–328.

Eckstein, Zvi, Suqin Ge, and Barbara Petrongolo. 2011. “Job and wage mobility with
minimum wages and imperfect compliance.” Journal of Applied Econometrics, 26(4): 580–612.

Engbom, Niklas, and Christian Moser. 2022. “Earnings inequality and the minimum wage:
Evidence from Brazil.” American Economic Review, 112(12): 3803–3847.

Hasumi, Ryo, and Tetsuaki Takano. 2024. “General Equilibrium Effects of the Minimum
Wage.” , (104). Available at SSRN: https://ssrn.com/abstract=4692738.

22
Hurst, Erik, Patrick J Kehoe, Elena Pastorino, and Thomas Winberry. 2022. “The
distributional impact of the minimum wage in the short and long run.” National Bureau of
Economic Research.

Manning, Alan. 2021. “The elusive employment effect of the minimum wage.” Journal of
Economic Perspectives, 35(1): 3–26.

Menzio, Guido, and Shouyong Shi. 2010. “Block recursive equilibria for stochastic models
of search on the job.” Journal of Economic Theory, 145(4): 1453–1494.

Menzio, Guido, and Shouyong Shi. 2011. “Efficient search on the job and the business
cycle.” Journal of Political Economy, 119(3): 468–510.

Neumark, David, and Olena Nizalova. 2007. “Minimum wage effects in the longer run.”
Journal of Human resources, 42(2): 435–452.

Rouwenhorst, K Geert. 1995. “Asset pricing implications of equilibrium business cycle mod-
els.” Frontiers of business cycle research, 1: 294–330.

Sabia, Joseph J, Richard V Burkhauser, and Benjamin Hansen. 2012. “Are the effects
of minimum wage increases always small? New evidence from a case study of New York state.”
Ilr Review, 65(2): 350–376.

Shi, Shouyong. 2009. “Directed search for equilibrium wage–tenure contracts.” Econometrica,
77(2): 561–584.

Statistics Bureau of Japan. 2024a. “2023 Yearly Average Results (Detailed Tabulation).”
https://www.stat.go.jp/english/data/roudou/results/annual/dt/index.html. Ac-
cessed: 2024-12-20.

Statistics Bureau of Japan. 2024b. “Labour Force Survey, 2023 Yearly Average Re-
sults.” https://www.stat.go.jp/english/data/roudou/results/annual/ft/index.html.
Accessed: 2024-10-19.

Tsuyuhara, Kunio. 2016. “Dynamic Contracts With Worker Mobility Via Directed On-The-
Job Search.” International Economic Review, 57(4): 1405–1424.

Zhao, Naibao, and Meng Sun. 2021. “Effects of minimum wage on workers’ on-the-job effort
and labor market outcomes.” Economic Modelling, 95: 453–461.

23