You are on page 1of 36

Extending a Career Choice Model: Two Regions and the Relative

Money Disinterest Bias

March 11, 2017

1
2
3
4
1 Introduction

2 The Model in Fonseca R., P. Lopez-Garcia and A. Pissarides

(2001)

The model in Fonseca R., P. Lopez-Garcia and A. Pissarides (2001) is an extension of the Mortensen-Pissarides

search and matching model (Dale Mortensen and Chris Pissarides, 1994; Pissarides, 2000). The extension

consists in the introduction of a career choice that each individual populating the economy faces, namely she

has to decide whether to be a worker or an entrepreneur. That choice is not only inuenced by the value of

the payos arising from the two paths but also by a random variable representing the entrepreneurial ability

of the individual and the number of jobs created by the rm she governs. Also, it is made the assumption

that the payo that comes from the career as entrepreneur increases constantly with the dimension of the

rm. Then, the choice of individual j is represented by the following expression:

j V U = j is an entrepreneur, otherwise she is a worker.

V and U are the usual Bellman equations that one nds in a Mortensen-Pissarides model.

Notice that being a random variable, it is possible to obtain the fraction of entrepreneurs in the population

through the pdf.

5
3 The Key Idea of the Thesis: The Relative-Money-Lightness Bias

In this section I will expose the extension to the Fonseca R., P. Lopez-Garcia and A. Pissarides (2001) I

would propose. The basic assumption is that if one considers a given population it may be that if one split

this population into some regions, the inhabitants belonging to region x may be endowed with a dierent

set of values, beliefs or more generally a dierent set of psychological characteristics with respect to region

y. Starting from this undeniable point I would go more in detail and propose a real life situation in which,

in my opinion, there exist two regions endowed with dierent relative weights attributed to the material and

non material goods they 'consume'. The population is the italian one and the regions I will look at are the

north and the south. The three 'goods' I would like to consider are 'money', 'family' and 'leisure' - as said I

will interpret the word 'good' in a broader sense, including both, say, materials and non material goods -. I

would assume that these variable can be approximately measured on a numerical scale and that there exists

some coecients - or weights - attached to each of them.

Just to be a little bit formal on this, the idea is the following:

UN (em , ef , ej ) = N m(em )0.5 + 1,N f (ef )0.5 2,N e, em + ef = e [0, E], E R+

US (em , ef , ej ) = S m(em )0.5 + 1,S f (ef )0.5 2,S e em + ef = e [0, E]

Where:

x(ex ) = ex f or x {m, f }

And the key assumption is:

N S
>
1,N + 2,N 1,S + 2,S

In what written above, ex stands for eort allocated to the activity x, N and S are 'north' and 'south'

respectively, m is money and f stands for family. Notice that m(em ) represents the utility that an individual

gains because of money. Then, I am not only talking about the satisfaction of being rich - as it may seem -

6
but also of the consumption of material goods, that is to say, of the material benets one can obtain from

money.

A simplifying assumption behind these formula is that an individual gains utility from two goods only apart

from the ones connected with money. Indeed, she gains utility from her family and loses utility from eort -

or equivalently gains utility from leisure -. Therefore I decided to consider the two mains non material goods

with respect to which I believe there exists such a 'relative lightness of money'.

Why eort and not time? Because I want to capture the fact that it is exactly the eort a possible source of

disutility and not the fact that we are allocating time to leisure. Then, one's utility is reduced if she makes

too much eort in making money or in constructing a good family.

Which is the scientic foundation of this idea? Other papers of Tabellini and other authors. Also, my

personal experience. I strongly believed that a well designed experiment would reveal this relative-lightness

of money.

To go more in detail on the issue, it is clear that in everyone's mind there exists two trade-os that one can

simplisticly dene 'money versus relax' and 'money versus family'. I will assume that the ratio between the

coecient regarding 'money' and the sum of the coecients on 'relax' and 'family' is higher in the north

with respect to the south. Notice that I am not referring to the trade-o 'job success' versus 'relax', in

what this trade-o should not exhibit such a great divergence in my opinion and according to the number of

southern individuals who reached very great performance in their job. Just to try to give a quite scientic

support to this last sentence, I will consider a job really appreciated for its social value and its responsabilities

namely the Contitutional Court's Judge. In this assembly the percentage of individuals belonging to the three

geographical areas of italy - north, center and south - is more or less the same. However, if one looks at the

last ranking of the italians reachest men drawn up by Forbes, we have that in the rst 10 positions there are

no individuals born in the south while the majority of them is born in the north.

Therefore, in the extension I propose, a southern individual gives more weight to a career as employee with

respect to a career as employer compared to a northern individual exactly because of the money-relax-family

7
issue I emphasized. Obviously a key assumption behind this choice is that the average amount of 'relax' and

'family' obtenable from a career as worker is higher with respect to the average amount of 'relax' and 'family'

obtenable from a career as entrepreneur, where I am talking about the average value with respect to time and

category of work. From a mathematical point of view, I assumed that a southern individual is aected by a

'relative-money-lightness bias' denoted by [1, ) that premultiplies w, where the lower is , the higher

is the bias. For a northern individual = 1, that is to say, she does not suer from any bias. Clearly I could

have decided to assume that the bias aects northern individuals, so that for example they suer from a sort

of 'relative-money-heaviness bias'. However, I decided to follow the logic of the world 'bias', that is to say,

a phenomenon that distinguish one or more agents inside a population, having negative consequences on it.

Since the results of the thesis will be that this bias determines more unemployment in the south, I decded to

attribute the bias to the southern individuals.

To conclude I would like to specify that phenomena like the lack of income mobility in the south contribute

to exacerbate this bias.

4 The Basic Model

4.1 The Model

The economy consists of a continuum of individuals in the unit interval with innite life span. I divide

the economy into two regions North and South. An individual belonging to South will be denoted by S,

while N N orth. S and N are exactly equal among them but for a parameter i , i {S, N } and for the

distribution from which they draw their level of 'entrepreneurial ability'. Indeed, each individual of group i

in the economy is endowed with an entreprenurial ability i , where:





N if i = N
i =

S
if i = S

8
with i [0, A], A <+ and i {S, N }. They are both exogenous parameters representing individual's

'entreprenurship', where their known distribution is F (.).

i represents also the number of jobs that individual i is able to manage if she chooses to become an

entrepreneur. Indeed, each individual can be either a worker or an entrepreneur and she makes her career

choice on the basis of the expected payos arising from the two scenarios. Entrepreneurs create jobs and

manage them; workers occupy them to make them productive. An entrepreneur can manage many jobs at

the same time but a worker can only occupy one job at a time. The wage is exogenous and constant, so that

it will be calibrated as a fraction of the production of the rm.

The allocation of jobs to workers takes place according to a process of search and matching, with the frictions

summarised in an aggregate matching function. Time is continuous and individuals are risk neutral.

The allocation of jobs is modelled as in Pissarides (2000, chapter 1), with the following modication. At

time t entrepreneurs manage n+v jobs, where n is the number of occupied individuals and v is the number

of vacancies. The number of individuals who decided to become workers is n+u where u is the number of

uneployed workers. Each occupied worker produces a constant ow of output denoted by y. At a constant

rate a shock arrives that interrupt the job relation so that the worker stops producing, becomes unemployed

and the entrepreneur opens another job to replace it.

The production technology is constant returns to scale, with labor as the only input.

The following details are referred to the steady state of the model:

Entrepreneurs and unemployed engage in a costly process of search and matching. During search, en-

trepreneurs suer a ow cost h for each vacant job and workers receive income b that can be interpreted

as an unemployment subsidy net of searching costs. The arrival of matches is determined by an aggregate

matching function with constant return to scale. As usual the arrivale process is summarized by the tightness

parameter i vi /ui such that: workers arrives to job at a Poisson rate q(i ), with elasticity in (1, 0), while

jobs arrive to unemployed according to a Poisson rate i q(i ) with eleasticity in (0, 1). Moreover:

9
lim q(i ) = lim i q(i ) = 0,
i i 0

lim q(i ) = lim i q(i ) =


i 0 i

Then, out of the steady state, the worker nd a job according to a Poisson process with time-varying arrival

rate i,t q(i,t ) and a vacancy is lled according to a Poisson process with time-varying arrival rate q(t ).

Moreover, a worker and a rm separate according to a Poisson process with arrival rate governed by the

stochastic separation rate t , leaving the worker unemployed and the rm with nothing. It is assumed that

either all the j vacancy of the rm are lled or they all remain empty. So I am using a Cobb-Douglas

matching function m(u, v) = u v 1 la Pissarides (2000), with matching rates q() for vacancies and

q() for the unemployed.

In what follows V i , J i , U i , Wi with i {S, N } denote respectively the present-discounted value of a vacant

job to an entrepeneur, of a lled job to an entrepreneur, of unemployment and employment for an agent

belonging to S or N.

rVi = h + q(i )[Ji Vi ]

rJi = yi wi + i [Vi Ji ]

rUi = b + i q(i )[Wi Ui ]

rWi = i wi + [Ui Wi ] , i {S, N }

On top of that employement at the regional level evolves according to

.
ni = [F (i ) ni ]i q(i ) i ni

So that after some rearrangement in steady state it holds:

i F (i )
ui =
i + i q(i )

As to S and

N , they will be dened soon.

The career choice of agent j in region i {S, N } is the following:

10
(j,i , i ,
, i )Vj Uj = j is an entrepreneur, otherwise she is a worker.

4.1.1 Some Details About (j,i , i ,


, i )

The function (j,i , i ,


, i ) represents the number of jobs created by rm j. Notice that ( . ) is not the

number of workers employed in rm j as it may be that the jobs created by the rm remains all vacant.

What I am saying is that either all the jobs in rm j are occupied or they are all vacant and this is true

by assumption. This assumption will be more clear in subsection 4.2. Notice also, that it is fundamental to

keep in mind that ( . ) is the number of jobs created by rm j and not the number of workers employed in

rm j, otherwise one cannot undestand the derivation of one of the equilibrium equation of the model.

The denition of ( . ) is the following for i {S, N }:

(j,i , i ,
, i ) j,i g(i ,
, i ) j i [|1 F (i ) [1 F (
)] |] , i [0, +)

Notice that ( . ) is, a function that depends not only on the level of entrepreneurial ability of individual j

as in the seminal paper, but also on the threshold value of entrepreneurial ability in region i, that is to say

i . i is indeed that value of j,i for which it holds [i g(i ,


, i )] Vi = Ui .

Therefore the number of jobs created by rm j decreases as the distance of 1 F (i ) from 1 F (
) increases.

That is to say, the number of jobs created is lower, the greater is the dierence between the fraction of the

population that chose to be entrepreneur in the region and the optimal fraction. I dened the optimal fraction

of entrepreneurs as 1 F (
) and, as I will specify in subsection 4.5, I set it at 0, 15. Behind this denition

there is the idea that the entrepreneurial activity is not only inuenced by the skills of the entrepreneur

but also on the sinergies that arise when there is a large fraction of the population that is composed by

entrepreneurs and by all the other secondary benets arising from various virtuousos phenomena connected

with having an optimal number of entrepreneurs.

From the form of ( . ), it is clear that a region with a number of entrepreneurs exactly equal to the optimal

one, creates j jobs. As in this region the number of entrepreneurs moves away from the optimal one, each

11
rm suers a loss, so that the number of jobs created is j,i i [|1 F (i ) [1 F (
)] |]. This loss arise

from the fact that in the regions, the rms are not exploiting entirely all the benets that come from having

an optimal number of entrepreneurs. So, as to i , it is a coecient that represents the magnitude of the

impact of this loss on the jobs created by rm j in region i. Notice that i varies depending on the region

we look at. I believe, indeed, that it is reasonable to assume that these sinergies are inuenced by a large

variety of variables that take on dierent values depending on the region we look at. So, as to this work, in

region N we have a more ecient economy, with a higher level of productivity - just to mention one variable

- with respect S. For this reason the loss due to the unexploitation of these sinergies is not so high with

respect to region S and N < S .

Also notice that for a large combination of calibrations it is necessary to assume that i exists and that it is

strictly higher than 1, if one believes that the sinergies inuence ( . ). Indeed, being F(.) a cdf, the value

of [|1 F (i ) [1 F (
)] |] is necessarily in between 0 and 1. Also in order to have a ( . ) close to the real

world number of jobs created by a rm, j has to be a number substantially greater than 1. Therefore, if

we want these sinergies to impact on the number of jobs created, we need a coecient that amplies the

distance between 1 F (i ) and 1 F (


).

To conclude, it is necessary to impose some constraints on the parameters and on j in order not to have the

counterintuitive result of a negative number of jobs created. After some computations reported in appendix

A.2, it comes out that the lower bound to impose to j,i is 6, 15 both for the north and for the south. There

are no upper bounds.

The reason why I need a new denition of jobs created by rm j with respect to the seminal paper, is that,

if I stuck to that denition then, I would obtain a result that contradicts the real world numbers. Indeed,

in the seminal paper there is no distinction between jobs created by rm j and entrepreneurial ability of j ,
since they are both denoted by . However, one of the scope of the work is to obtain that the fraction of

entrepreneurs in the population is higher in the north with respect to the south, but this holds if and only if


1 F (N ) > 1 F (S ), which happens if and only if S > N

, so that the average number of jobs created
in region N would result higher with respect to region S which is not true in the real world.

12
4.2 Microeconomic Basis of the Model

Since I did not nd any well specied mathematical formulation of the objects that represent the core of the

model, I decided to write them carefully.

In what follows will use the expression action not because we are in a game theoretical framework with

strathegic thinking but because I nd this expression more salient. Indeed, individuals not only choose their

consumption, but make also qualitative choices such as the career choice. I will expose the formulation as if

time is discrete, since I nd this formulation more clear. In what follows I will neglect the discounting rate

in order to simplify the exposition.

The set of agents is denoted by J, where J is a continuum of individuals with dimension normalized to one.

As said J = {South, N orth}. A representative individual belonging to the former group is called S, while

N is the representative individual belonging to N orth. In what follows ( . ) is the number of jobs created

in that specic rm. A detailed specication of it will be done in the next section.

The action spaces are for t = 1, 2, 3, ...:

Atj = ACareerChoice AU nemployed AtEmployee AStartupper AtEntrepreneur , j {S, N }

ACareerChoice = {W orker, Entrepreneur}

AU nemployed = ( , b ]

AtEmployee = ( , wt ]

AStartupper = ( , ( . )h ]

AtM anager = ( , ( . )(y wt ) ]

So that atj Atj , atj,CareerChoice ACareerChoice , atj,U nem AU nemployed , atj,Empl AtEmployee , atj,Start

AStartupper and atj,M an AtM anager .

13
We need to specify also the set of the status and the existence of a consumption function that will be dened

later:

Statust {U nemployed, Employee, Startupper, M anager, T o be chosen}

ctj : Atj <.

Notice that for some action subspace the superscript t is not needed in what they do not depend on time,

while they are always needed for the choice variables, in what the choices happen at any period.

I used the expression Employee instead of worker, in order not to make confusion in the context in which the

individual chooses to be a worker, and in the successive instant she is unemployed. For the same reason I

denoted by Manager the individual who chooses to be an entrepreneur and who faced a match in her vacant

jobs. Note that the model is based on the simplifying assumption that either all the jobs of a given rm

are occupied or they are all empty, so that when the individual is a Manager, she can choose to consume

( . )(y w). Note also that the denition of Atj requires that at each time the individuals chooses an action

in all of the dened subsets ACareerChoice , AU nemployed , AtEmployee , AStartupper and AtM anager . As will be

clear when I will dene the utility function, this does not create any unrealistic mechansim. Indeed when,

for example, the individual chooses to be a worker, her utility is not inuenced by the choice made in the

subset AStartupper .

For t = 1, 2, 3, ... the expected utility function is:

Et uj (atj , Statust ) + uj (at+1


 
j , Statust+1 ) + ... =

h i

t
P t+i t+i
a + p(U nem |U nem )a + p(Empl |U nem ) a if Statust = U nemployed




j,U nem t+i t j,U nem t+i t j j,Empl
i=1




h

i

t
p(U nemt+i |Emplt )at+i t+i
P
j,U nem + p(Emplt+i |Emplt )j aj,Empl

j aj,Empl + if Statust = Employee




= i=1
h i
t+i t+i

t
P
a + p(Start |Start )a + p(M an |Start )a if Statust = Startupper


t+i t t+i t

j,Start

j,Start j,M an


i=1


h i

t
p(Startt+i |M ant )at+i t+i
P
aj,M an + j,Start + p(M ant+i |M ant )aj,M an if Statust = M anager




i=1

where for s = 1, 2, 3, ... the rigorous denition of uj (.) is:

14

asj,U nem




if Statuss = U nemployed



as


j j,Empl if Statuss = Employee
uj (csj (asj , Statuss )) =
asj,Start




if Statuss = Startupper



asj,M an

if Statuss = M anager

where csj represents the amount of consumption at time s and as implicitly said uj (csj ) = csj and:


asj,U nem




if Statuss = U nemployed



as


j j,Empl if Statuss = Employee
csj (asj , Statuss )) =
asj,Start




if Statuss = Startupper



asj,M an

if Statuss = M anager

So that I simply wrote uj (atj , Statust ) in order not to complicate the notation.

We need to specify also that:

asj Asj , , j {S, N }

and the following state equation holds:

Statuss = f (Statuss1 , a1j,CareerChoice ) =







U nem. if Statuss1 =U nem., w.p. 1s q(s ) or Statuss1 =Empl., w.p. s or a1j,CareerChoice =W orker and s=1





Empl. if Statuss1 =U nem., w.p. s q(s ) or Statuss1 =Empl., w.p. 1s
=


Start. if Statuss1 =Start., w.p. 1q(s ) or Statuss1 =M an., w.p. s or a1j,CareerChoice =Entrepr. and s=1








M an. if Statuss1 =Start., w.p. q(s ) or Statuss1 =M an., w.p. 1s

with Status0 = T o be chosen.

Note that the above probabilities are written in a simplied notation in what a rigorous one would require

to write, for example, p(States+i = U nem|States = U nem) instead of p(U nems+i |U nems ), moreover we

assumed that the collection of t is constant over time, so that we can neglect the subscript.

15
Also, the probability that the individual is unemployed at s + i, given that she was uneployed at s is the sum

of the probabilities expressing all the possible routes through which this could happen, i.e:

p(U nems+i |U nems ) = p(U, U, ..., U )+p(U, E, U, ..., U )+p(U, U, E, U, ..., U )+...+p(U, U, ...U, E, U )+p(U, E, E, U, ..., U )+

+p(U, U, E, E, U, ..., U ) + ... + p(U, ..., U, E, E, U ) + ... + p(U, E, ...E, U )

Where for example p(U, U, E, U, ..., U ) is the probability that the individual is unemployed at time s, then

again unemployed, then employee, then again unemployed up to time t+i. This probability can be constructed

in a straightforward way starting from the following transition probabilities:

prob(U nems |U nems1 ) = 1 s q(s )

prob(Empls |U nems1 ) = s q(s )

prob(U nems |Empls1 ) =

prob(Empls |Empls1 ) = 1

prob(Starts |Starts1 ) = 1 q(s )

prob(Starts |Ents1 ) =

prob(Ents |Starts1 ) = q(s )

prob(Ents |Ents1 ) = 1

Obviously the model is based on the simplifying assumption that the probability that an employee becomes

unemployed is the same as the probability that an entrepreneur looses all of her job relations contemporarly.

Adjusting this assumption does not change the results reported in the text.

Therefore, at period 1 we have the expected utility from which we obtain the equilibrium condition
1 Vtot

j V = Uj :
1 Notice that Vtot is the present expected utility of choosing Entrepreneur obtainable from the following expression. In order
to highlight the role of j one can simply bring it outside Vtot .

16
 
E1 uj (c1 (a1j , Status1 )) + u(c2 (a2j , Status2 )) + ... =

h i
p(U nem2+i |U nem2 )a2+i 2+i

a2j,U nem + if a1CareerChoice = W orker
P
+ p(Empl |U nem ) a

2+i 2 j j,Empl

j,U nem
i=1
= h i
a2j,Start + p(Start2+i |Start2 )a2+i 2+i
if a1CareerChoice = Entrepreneur
P

j,Start + p(M an 2+i |Start 2 )aj,M an
i=1

Notice indeed that the state equation is constructed in way such that the individual is 'Unemployed' at period

2 if chose 'Worker' at period 1 and as 'Startupper' if chose 'Entrepreneur'.

4.3 The Equilibrium Equations

In this subsection I will expose the equilibrium equations of the model. The step-by-step derivation is in the

appendix as the scope of this subsection is to give an intuitive explenation of the relations that govern the

model.

Combining [i g(i ,
i , i )] Vi = Ui with the Bellman equations exposed at subsection 4.1, I obtain the

Entrepreneurship Curve. It is called in this way because

The second equilibrium equation is the Job Creation Curve. Through

i
A
x dF (x)
i i + i q(i )
= i
g(i ,
i , i )
i + i q(i ) F (i )

To conclude I have the third equilibrium equation, that is to say, the Wage Curve. The Wage Curve

Therefore the equilibrium equations for region i {N, S} are - the step by step derivations are in the appendix

-:

q(i )(yi i wi ) hi (ri + i ) i q(i )i wi + b(ri + i )


[EC] [i g(i ,
i , i )] =
(ri + q(i ))(ri + i ) q(i )i (ri + i q(i ))(ri + i ) i q(i )i

17
A
i i + i q(i )
x dF (x) [1 F (i )] g(i ,
, i )
i
[JCC] =
i + i q(i ) F (i )

i i (yi + h)(ri + i + i q(i )) + (1 i )b(ri + i + q(i ))


[W C] wi =
i i (ri + i + i q(i )) + (1 i )i (ri + i + q(i ))

4.4 Uniqueness

In this subsection I will prove the uniqueness of the results that will be shown in the next one.

To be continued...

4.5 Calibration

In this paragraph I will calibrate the model. First of I have to specify that the probability distribution

function that describes the distribution of the entrepreurial ability in the two populations is a triangular one.

For this reason in the following table the symbols A and c appear. The functions are indeed dened on the

following domain: [0, A]. As to c it is the value of for which the pdf reaches its maximum.

18
In what follows I will justify this calibration .
2

2 To make more clear the notation I neglected the subscript that refers to the region when this does not create confusion.

19
As to , I computed it in the following way:

Separations11:311:4 U nemployed11:4 U nemployed11:3 + (Employed11:4 Employed11:3 )


= u
Employed11:3 Employed11:3

Where in writing x:y I mean quarter y of year x. As said, I chose to calibrate the model looking at this

period because it is the last period in which it is possible to have data about a key variable of the work, that

is to say number of rms in region i. As said again, the latter is fundamental since through it I can compute

the fraction of individuals who chose to be entrepreneurs. The above formula is an approximation since I

neglected the (almost irrelevant) impact on Separations, of those ows inherent the inactive, the pensioners

and the individuals entering in the labor force.

As to r, I set it at 0, 01 since I am calibrating the model on a quarterly base, so that 0, 01 comes out from

the standard assumption of a risk free annual real interest rate of 5%.

As to it is the exponent of q() = . Consistently with most of the literature (Petrongolo and Pissarides,

2001), the value of is equal to that of , that is to say 0, 5.

As to , the choice has beend done with the scope of getting results as close as possible to the real world

ones as usually done in many papers. The choice of A has been done also on the basis of the numbers that

the model gave me. As to c it is equal to 0, 1. The only thing I would like to specify here, is that, the only

constraint I decided to impose to c before setting it, is to have a distribution in which it is very low the

number of individuals with a high level of entrepreneurial ability. I did so, since I nd this assumption an

intuitive one. In numerical terms this means that I prefered to set a value for c not greater than 1% of A.

The value of h has been established in an intuitive way also, since I do not believe it is possible to nd

real world numbers able to support a calibration of it. In the work h is indeed the cost of keeping a vacant

position open. My choice has been to set this value as 5% of yi . As to the latter I normalized to 1 the number

expressing the value added per occupied in 2011 in the north. That way I observed the same variable for the

south and it appeared to be the 76% of the one of the north.

However, all the calibrations have been clearly done separately for each region.

20
Let us now move to b. In 2011 the rule regarding the unemployment benet was that its value had to be

equal to the 75% of the wage for the rst 6 months, then 50% for the successive 6 months and then equal

to the 35% of the wage. The data show an average unemployment spell of 2 years, therefore it comes out a

value of b that is 0, 49 times the wage. As to the latter, I computed it as follows:

salaries
w = y. More details about the computation of w in
value added to the cost of the f actors of production
chapter 3.6.

Let us now move to 1 F (


). As said the fraction of entrepreneurs in the north is around 4%, while for

the south it is more or less 2%. I do not believe that it is reasonable to say that this averages represents

desirable values, for the simple reason that these two regions are not ones in which the economic system

works with a very high level of eciency. So, I decided to look at an Italian city that should be one of the

more ecient one, at least from an economical point of view. This city is Milan. Taking into account the

entire population of Milan, the fraction of entrepreneurs comes out to be around 8%. It is clear however that

this value incorporates the suburbs of the city so that 8% cannot represents the optimum I was looking for.

Therefore, I decided to choose a number higher than 8% and taking also into account how the results of the

model reacted to the dierent choices I decided to set it at 15%.

As to the calibration has been inuenced by how the model reacted to dierent values of this variable. The

only imposed constraint - apart from the trivial one i 0 - is N < S since - as explained in section ? -

region N is more ecient than region S, so that the loss due to the unexploitation of the benets of having

a number of entrepreneurs close to the optimal one, is lower in the north with respect to the south.

To conclude we are left with S . It should be necessary to have access to a sperimental background in order

to calibrate this parameter. However, the scope of this work is not to obtain numerical results that exactly

correspond to the real world numbers, but just to claim that, if someone were able to prove the existence of

the RMLB than this work may describe a possible model to study its consequences on the relevant variables

of the labor market. Therefore, I calibrate S with an euristic/intuitive way, in the sense that I calibrate

it by trying to be consistent with some rules that I nd reasonable or at least intuitive. For example I do

not believe it is reasonable to claim that a southern individual would overvalue a arising from a career
21
as employee more than 40% with respect to a northern one. At the same time I do believe there exists this

bias. Then looking at how the results of the model reacted to dierent calibrations I decided to set S equal

to 1, 3.

4.6 Results

Result 1 The level of entrepreneurial ability above which an individual chooses to be entrepreneur is higher

in the south with respect to the north. That is to say, the minimum level of entrepreneurial ability required

to be entrepreneur is higher in the south with respect to the north. Then, the fraction of individuals who

choose to become entrepreneur is higher in the north with respect to the south.

Proof: It is necessary to apply the following code on MatLab in order to obtain the equilibrium values:

Where TesiN is the name I gave to the equilibrium equations that hold for the north.

The results are:

N = 1, 44


N = 6, 71

wN = 0, 69

As to the south the code is:

22
Where TesiS is the name I gave to the equilibrium equations that hold for the south.

And the results are:

S = 0, 21

S = 6, 77

wS = 0, 22

So that it holds:

0, 10 = 1 F (S ) < 1 F (N

) = 0, 11 @

Result 2 The average number of jobs created per rm is higher in the north with respect to the south.

Proof: As said the denition of average entrepreneurial skill level in region i is:

A
xf (x)dx

E[i |i i ] = i

1 F (i )

While the average number of jobs created per rm in region i is:

E [( . )] = E[ji |ji i ] i [|1 F (i ) (1 F (i ))|]

As to the north it holds:

A
xf (x)dx

N
E[N |N N ] = )
1 F (N

0, 34 103 10 6, 712 + 0, 66 6, 713


=
10 (10 0, 1) (2 10 6, 71 10 0, 1 6, 712 )

89, 15
= = 8, 24
10, 82

23
So that:

E [( . ) | i = N ] = 8, 24 110 [0, 15 0, 11] = 3, 84

As to the south:

A
xf (x)dx

E[S |S S ] = S

1 F (S )

0, 34 103 10 6, 772 + 0, 66 6, 773


=
10 (10 0, 1) (2 10 6, 77 10 0, 1 6, 772 )
86, 46
= = 8, 29
10, 43

So that:

E [( . ) | i = S] = 8, 29 140 [0, 15 0, 10] = 1, 29

Result 3 Unemployment is higher in the south with respect to the north.

Proof: As said unemployment in region i is:

i F (i )
ui =
i + i q(i )

So that :
3

0, 01 0, 89
uN, = = 1, 46%
0, 01 + 0, 6
0, 01 0, 90
uS, = = 4, 74% @
0, 01 + 0, 18

Result 4 If the Relative-Money-Lightness Bias did not exist, unemployment and the threshold value of

would be lower in the south, while salaries, tightness, the number of entrepreneurs and the average number

of jobs created per rm, would be higher.

3I added the subscript to distringuish these unemployment rates with the one found in 'Result 4', in which it is assumed
that the RMLB does not exist.

24
Proof: Setting S = 1, and applying the mentioned code used for TesiS, one gets:

S = 0, 42

S = 6, 75

wS = 0, 32

That implies:

1 F (S ) = 0.107

Also:

A
xf (x)dx

E[S |S S ] = S

1 F (S )

0, 34 103 10 6, 752 + 0, 66 6, 753


=
10 (10 0, 1) (2 10 6, 76 10 0, 1 6, 762 )

= =

So that:

E [( . ) | i = S] = 140 [0, 15 0, 107] = @

And:

0, 01 0, 893
uS = = %@
0, 01+

4.7 Real World Numbers: Comparisons and Derivations

In the following table I will compare the numerical results of the work with the real world numbers. As

said the scope of the work is not to obtain numbers equal to those of the real world, but to reproduce the

25
'qualitative results' one can nd in the real world. By 'qualitative results' I mainly mean relations among

variables, in the sense that one variable takes on a numerical value higher/lower with respect to another

one. Notice that this objective has always been reached. At the same time, I tried to reach the objective of

'not going too far' from the real world numbers, where by 'not too far' I mean no more/less than two times

higher/lower the values observable in the real world. I failed this objactive for two variables only, that is to

say with the fraction of entrepreneurs in the population and with the unemployment. With both variables

however the obtained numbers are no more/less than four times the real wolrd values.

The values obtained for 1 F (i ) are more or less four times the ones observable in the real world. This will

impact on the results about the unemployment, since u decreases as 1 F ( ) increases. As to the distance

among

1 F (N ) and 1 F (S ), it is twice the correct one. As to all the other variables they are always

less/more than two times higher/lower with respect to the real world values, save for N q(N ) and wN which

are less than three times lower than the correct numbers.

26
In what follows I will justify the values obtained for the real world.

I used as a measure for the fraction of the population that chose to be entrepreneur for each region the

following one :
4

# f irms with more than 2 employee


=
# italian inhabitants with age [18, 60]

I searched for these data in the standard dataset of the Italian Institute of Statistics (ISTAT), but I did not

nd the number of rms with more than two employee in each region. For this reason I searched for this

5
value elsewhere and I found them in the Census For Industries and Services , that incorporated the data in

a dataset which is dierent with respect to the standard one that ISTAT holds, although it is also managed

by ISTAT itself. However, that Census reports the data I looked at, only up to 2011, so that I have been

forced to calibrate the entire model on that year. As to the value for the number of italian inhabitants with

age between 18 and 60 per region I easily found this datum in the standard dataset of ISTAT. The reason

why I chose to look at rms with more than two employee only is that at the basis of my idea there is the

assumption that a southern individual overestimate a certain variety of goods when she compares it with

money with respect to a northern guy. For this reason in her career choice a southern individual evaluate in

a dierent way - with respect to a northern one - not a career as head of a family business but to a career as

owner of a rm with a level of prot at least above the average one in the region. However, it is not simple

to disantagle among these two career paths, so that I decided to label the second one as a one in which the

entrepreneur manage a rm with (strictly) more than two employee.

As to the denominator, I looked at the italian inhabitants between 18 and 60 years old because the RLMB

distinguishes the individuals that have grown up in the south of Italy with respect to those who grown up in

the north, so that it would have made no sense to look at non italian individuals. Clearly, for this reason, it

would have been better to have at the numerator the number of rms with more than two employee held by

italians only. However, I have not been able to nd these data and I believe they cannot be found. As to the

4 In order not to make the notation too obscure I neglected the subscripts i and 2011. However, I computed for both regions
in 2011.
5 In particular I looked at the following website http://dati-censimentoindustriaeservizi.istat.it/Index.aspx#

27
age range, I chose it because I believe it is the one that better captures those individuals that make choices

inherent their career.

It comes out that S = 2, 19% while N = 4, 19% . The region chosen for the south are: Molise, Campania,

Puglia, Calabria and Sicilia. I have been forced to exclude Basilicata since there is clearly an error in the

value representing the number of rm, since it comes out that Basilicata is equal to 22, 91%. As to the

north I looked at the following regions: Piemonte, Valle d'Aosta, Lombardia, Trentino-Alto Adige, Veneto,

Friuli-Venezia Giulia, Liguria and Emilia Romagna. I did not choose deliberately this grouping but I used

the usual one established by ISTAT.

The second variable is . There are clearly no problems in nding data about the number of unemployed

individuals. As to the number of vacancies I looked at the website fece4job.com. I have not been able to

nd data for 2011, as the last one that can be found refers to 2013 and it says that the number of vacancies

in Italy are more or less 1.300.000. The site clearly have an incentive to come out with a too large number,

nevertheless the algorithm used tries to discriminate among true and false jobs openings. Also, it is necessary

to say that it looks at online announces only so that there exists an underestimation eect also. Assuming

these two eects more or less compensate each other, I decided to use the above mentioned database. I would

like also to specify that ISTAT provides a statistic called 'jobs vacancy ratio' that is obtained through the

V
following formula j.v.r. = V +E where E stands for employed. However, the jvr has a problem since it is

computed on the base of numbers regarding those rms that have more than ten employee. This implies

that the jvr is substantially lower with respect to the one of the entire economy as a great variety of small

rm with a low number of workers employed are neglected. Indeed, according to ISTAT, Italy should have

a number of vacancies that amounts aprroximatively to 100.000. Since I have not been able to nd data at

a regional level, I used the information provided in a newspaper that extrapolated the data - now deleted

- from face4job.com. According to that newspaper Piemonte has 78.200 vacancies while campania 77.300.

As to other regions there are no information. Therefore I decided to approximate the value of VS and VN
V acancies
by assuming that in each region of the north there is the same ratio ratio = . I assumed the
Inhabitants
78.200
same for the south. That way the ratio for a northern region is equal to ratioN = = 1, 78%
4.403.000
77.3000
while ratioS = 5.851.000 = 1, 32%. Multiplying ratioN by the number of inhabitants of each region in the

28
north and summing the obtained numbers I got VN = 492.875, while for the south I obtained VS = 235.969.

Therefore it comes out that N = 0, 69 while S = 0, 34. The dierence among the two 's is not as great

as the dierence among the two V 's would suggest since for the north I considered an area with 27.751.000

inhabitants while for the south they are 17.858.000. That way the total number of vacancies for the north is

clearly higher with respect to the one of the south.

M atchings11:311:4
The third variable is q(). I computed it with the following formula: q() = where I
U nemployed11:3
approximated the number of matching as:

M atchings11:311:4 u Employed11:4 Employed11:3 + (Retired11:4 Retired11:3 ) + Separations11:411:3

It is an approximation since I neglected the impact of the ows inherent the inactives and the new individuals

entering in the category labor force. However, this should not distort that much the numbers.

As to the number of separations I already computed it when I calibrated the variable , so that, the derivation

of it can be found in section ?.

The description of the derivation of q() can clearly be neglected as all the numbers necessary to compute it

all already been found. As to w, I computed it as a fraction of the productivity level. The computation of the
salaries
fraction I applied to yi is the following, f raction = . Both
value added to the cost of the f actors of production
the numerator and the denominator are numbers collected by ISTAT - at a regional level - that corresponds

to the sum of all the salaries paid in a given region for the former. The same method is used for the latter.

The obtained values are


6 f ractionN = 0, 39, f ractionS = 0, 38.

Now we are left with the last variable, the average number of jobs created per rm. Since many of the

previous variables refer to italians only and since the scope of the work is to analyse the behaviors of the

italians, it would have been better to comput this value for the italian workers only. Also, since I looked at

rms with more than two employee in computing the fraction of the population composed by entrepreneurs,

it would have been desirable to loke at those rms with more than two employee only. However, notice that

the formula to compute the average number of jobs created per rm in the real wolrd is:
6 For a detailed computation go to appendix A.2

29
# of workers + vacancies
a.n.j.c.
# f irms

So, in order to satisfy the two above mentioned requirements I would need the number of vacancies in those

rms with more than two employee and the number of italian workers in those rms. I have not been able

to nd these data. For this reason I have been forced to look at whatever rm existing in region i. The

obtained value are represented in the table, that is to say, 3, 73 for the north and 1, 83 for the south.

5 Conclusions

Appendix

A.1 Derivation of the Equilibrium Equations

A.1.1 Entrepreneurship Curve

The Entrepreneurship Curve can be obtained starting from:

[1] i ( . )Vi = Ui

Where i is the number of jobs created by rm i, and has been dened as:

i ( . ) [i g(i ,
)]

So that [1] becomes:

[1.2] [i g(i ,
, i )] Vi = Ui

Where we sobstitute Vi and Ui from the well known Bellman equations:

30
[2] rVi = hi + q(i ) [Ji Vi ]

hi + q(i )Ji
Vi =
r + q(i )

[3] rJi = yi wi + i [Vi Ji ]

yi wi + i Vi
Ji =
r + i

[4] rUi = bi + i q(i ) [Wi Ui ]

bi + i q(i )Wi
Ui =
r + i q(i )

[5] rWi = i wi + i [Ui Wi ]

i wi + i Ui
Wi =
r + i

So that plugging [3] into [2] and [5] into [4] I get respectively:

yi w i + i V i
hi + q(i )
r + i
[2.2] Vi =
r + q(i )
yi wi
hi + q(i )
r + i
Vi =
i
r + q(i ) q(i )
r + i
hi (r + i ) + q(i )(yi wi )
Vi =
(r + q(i ))(r + i ) i q(i )
i w i + i U i
bi + i q(i )
r + i
[4.2] Ui =
r + i q(i )
i wi
bi + i q(i )
r + i
Ui =
i
r + i q(i ) i q(i )
r + i
bi (r + i ) + i q(i )i wi
Ui =
(r + i q(i ))(r + i ) i i q(i )

Now, I can plug [2.2] and [4.2] into [1.2]:

hi (r + i ) + q(i )(yi wi ) bi (r + i ) + i q(i )i wi


[i g(i ,
, i )] = [EC]
(r + q(i ))(r + i ) i q(i ) (r + i q(i ))(r + i ) i i q(i )

31
A.1.2 Job Creation Curve

The Job Creation Curve can be obtained starting from the intuitive relations:

.
[1] ni = ui i q(i ) i ni

.
[2] ni = vi q(i ) i ni

Where the number of vacancies in region i, is simply the dierence between the number of entrepreneurs

times the average number of jobs created per rm and the number of jobs occupied:

[3] vi = [1 F (i )] E [i ( . )] ni

This formula can be simplied as it holds:

E [i ( . )] = E [i g(i ,
) | i i ]

= E [i | i i ] g(i ,
)

A
x dF (x)

= i
g(i ,
)
1 F (i )

So that [3] becomes:

A
[3.1] vi = x dF (x) [1 F (i )] g(i ,
)

i

As to the number of uneployed, they are simply the dierence between the number of individuals who chose

to be workers and the number of jobs occupied:

[4] ui = F (i ) ni

So that [1] and [2] become:

.
[1.2] ni = [F (i ) ni ] i q(i ) i ni

32
A
" #
.
[2.2] ni = x dF (x) [1 F (i )] g(i ,
) ni q(i ) i ni

i

That imply:

A
" #
[5] [F (i ) ni ] i q(i ) i ni = x dF (x) [1 F (i )] g(i ,
) ni q(i ) i ni

i

A
" #
[F (i ) ni ] i = x dF (x) [1 F (i )] g(i ,
) ni

i

Moreover, [1.2] in steady state implies:

F (i )i q(i )
[1.3] ni =
i q(i ) + i

While [2.2] in steady state implies:

A
" #
x dF (x) [1 F (i )] g(i ,
) q(i )

i
[2.3] ni =
q(i ) + i

So that plugging [1.3] and [2.3] into [5] I get:

A
" #
x dF (x) [1 F (i )] g(i ,
) q(i )
A
" #
F (i )i q(i )
 

i
F (i ) i = x dF (x) [1 F (i )] g(i ,
)
i q(i ) + i i
q(i ) + i


2
  "A # 
i q( i ) q(i )
F (i ) i = x dF (x) [1 F (i )] g(i ,
) 1
i q(i ) + i
i
q(i ) + i

A
x dF (x) [1 F (i )] g(i ,
)
i [i q(i ) + i i q(i )] i + q(i )
i
=
i q(i ) + i i + q(i ) q(i ) F (i )

A
x dF (x) [1 F (i )] g(i ,
)
i 2i+ i i q(i )
i
=
i q(i )i + 2i F (i )

A
x dF (x) [1 F (i )] g(i ,
)
i i + i q(i )
i
= [JCC]
i q(i ) + i F (i )

33
A.1.3 Wage Curve

As usual, the equilibrium value of wi can be otbtained through the formula:

max(Ji (wi ) Vi )1i (Wi (wi ) Ui )i


wi

Where the FOC will be slightly dierent with respect to the standard one because of i . Indeed, we have:

0 0
F OC : (Wi (wi ) Ui )i (1 i )(Ji (wi ) Vi )i Ji (wi ) + (J(wi ) Vi )1i i (Wi (wi ) Ui )i 1 Wi (wi ) = 0

1 i
(Wi (wi ) Ui )(1 i ) = (Ji (wi ) Vi )i
ri + i ri + i

(Wi (wi ) Ui )(1 i ) = (Ji (wi ) Vi )i i

Where now we can sobstitute Wi (wi ) Ui and Ji (wi ) Vi in order to isolate wi :

i wi b yi wi + h
(1 i ) = i i
ri + i + i q(i ) ri + i + q(i )

(1 i )(i wi b)(ri + i + q(i )) = i i (yi wi + h)(ri + i + i q(i ))

wi [(1 i )i (ri + i + q(i ) + i i (ri + i + i q(i ))] = i i (yi + h)(ri + i + i q(i )) + (1 i )b(ri +

i + q(i ))

i i (yi + h)(ri + i + i q(i )) + (1 i )b(ri + i + q(i ))


wi = [W C]
i i (ri + i + i q(i )) + (1 i )i (ri + i + q(i ))

A.2 The Constraints on S and N

To derive the constraints on S and N , we need to impose that the rm with the lower number of jobs

created - in region i - has a ( . ) higher than 0. Notice that the rm with the lower number of jobs created

is also the rm with the lower level of entreprenurial ability in region i, that is to say the rm for which it

holds j = i . This is true because the denition of ( . ) is:

34
(j , i ,
, i )

So, ( . ) is higher than 0 if and only if:

j i [|1 F (j ) (1 F (
))|] 0

Where j is distributed as a triangular. So, according to the calibration of the model it becomes:

" #
100 + j2 20j
j i | 0, 15| 0
100 10

When 1 F (j ) is strictly lower than 0, 15 - which means that j is strictly higher than 6, 15 - the inequality

becomes:

100+2j 20j
h i
j i 0, 15 10010 0

 
j i 0, 15 1, 11 + 0, 01j2 0, 22j 0

So that, for region N we have the constraint N 26.36 which is always satised being i [0, 10] with

i {S, N }. As to the negative constraint it is also satised for the same reason. As to region S the constraint

is S 26, 31 and for the same reason said above we have no additional constraints on S .

In case 1 F (j ) is strictly higher than 0, 15 - which means that j is strictly lower than 6, 15 - the inequality

becomes:

 
j i 1, 11 0, 01j2 + 0, 22j 0, 15 > 0

So that for region N we have the constraint N 25, 11, so that it is never satised. As to region S the

constraint is S 25, 15 and again the constraint is never satised. As to the negative solutions of the

inequalities they are always satised since we are in the case in which j is strictly lower than 6, 15.

As to the case in which j is equal to 6, 15 the inequality is trivially satised.

Summing up, the constraints are:

i 6, 15 , i {S, N }

35
A.3 Details About the Real World Numbers

SHOULD I INCLUDE THE 'NOT VERY IMPORTANT' NUMBERS USED IN THE EXCEL FILES, SUCH

AS THE SALARIES PAID IN REGION X, THE SALARIES PAID IN REGION Y.... SO THAT I TOTALLY

JUSTIFY THE OBTAINED NUMBERS?

salaries
As said f raction = . So that:
value added to the cost of the f actors of production

177.120.494.000
f ractionN = = 0, 39
457.843.942.000
38.894.293.000
f ractionS = = 0, 38.
101.668.364.000

Then:

wN = f ractionN yN = 0, 39 1 = 0, 39

wS = f ractionS yS = 0, 38 0, 76 = 0, 29.

References

Websites

http://www.linkiesta.it/it/article/2014/12/01/la-disoccupazione-e-il-falso-mito-dei-posti-vacanti/23700/

http://dati-censimentoindustriaeservizi.istat.it/Index.aspx#

36

You might also like