LABOUR ECONOMICS
Les 1: Introduction to labour economics
What is labour economics?
Why do we care about labour markets?
- Work makes an important part of each individuals life
o Income generating, that determines consumption
o Together with education, health, …
§ If you don’t work you don’t have money, you can’t consume anything,
you have poor health and education
- Labour is key for production
o Influences innovation and economic growth
- The labour market is subject of many social policies
o bv. social security, immigration, …
Who are the actors of the labour market?
Firms
= labour demand; how many workers to hire, wages, prices to set, invest in technology?
o maximise profits
o constraints
§ labour and capital = limited inputs
§ competition and consumer demand holds back demand
Workers
= labour supply; how many hours to work, effort at work, skills, ..
o maximise their well-being
§ utility from leisure, consumption, health, …
o constraints
§ hours in a day = limited
§ income and job offers
Government
= rules of the labour market
o maximise social welfare
o constraints
§ tax income
§ elections
o their own policy tools
§ minimum wage, health and safety regulations, immigration, taxation
via income, payroll
, 4 Chapter 1
FIGURE 1-1 Supply and Demand in the Engineering Labor Ma
The labor supply curve gives the number of persons willing to supply th
wage. The labor demand curve gives the number of engineers that firms
where supply equals demand, so that 20,000 engineers are hired at a wa
Labour supply curve: upwards sloping, Earnings ($)
workers are willing to work more if they Labor Supply
receive higher wages. Curve
50,000
Labour demand curve: downwards sloping,
firms are willing to hire more workers if they
can offer lower wages. Equilibrium
40,000
è No government intervention: the
intersection of the supply and demand Labor Demand
Curve
curves (only) determines market clearing 30,000
wages and equilibrium.
Employment
10,000 20,000 30,000
How do we study labour markets? when labor is expensive. The relation between
DATASOURCES firms are willing to hire is summarized by th
in Figure 1-1. As drawn, the labor demand c
industry want to hire 20,000 engineers when th
- Cross-sectional household surveys
engineers if the wage rises to $50,000.
o Current population survey (US)
Workers and firms, therefore, enter the lab
o European labor force survey workers are willing to supply their services
o Census willing to hire them. Conversely, few worker
the wage is low, but many firms are looking f
- Longitudinal household surveys firms search for workers, these conflicting des
o Panel study of income dynamics reaches an equilibrium. In a free-market eco
o British cohort study equals demand.
o German socio-economic panel As drawn in Figure 1-1, the equilibrium wa
hired in the labor market. This wage–employm
- Administrative data it balances out the conflicting desires of wor
o Taxation data the engineering wage was $50,000—above e
o Social registry data only 10,000 engineers, even though 30,000 e
number of job applicants would bid down th
available. Suppose, instead, that the wage w
1) Labour economic theory engineers are cheap, firms want to hire 30,00
willing to
= Simplification of the real world designed to isolate the core mechanism work at
behind onethat wage. As firms compe
specific
question. up the wage.
There is one last major player in the labor
can tax the worker’s earnings, subsidize the tra
We will make simplifying assumptions, objective function and their constraints, optimal
behavior, comparative statics (‘all else equal’)
Example: slide 14
bor04724_ch01_001-018.indd 4
, 2) Labour economic empirics
Descriptive analysis
= establish facts about economic patterns we observe in the real world, provide basis for
theory
Causal analysis
= test theoretical predictions with real world data and establish causal relationships. You
need to look for the effect of something happening with all other things equal.
! Problems with examine an effective causal relationship within the observational data:
- Endogeneity: X reacts on Y
- Omitted variable bias: Z causes effect on Y and not X
- Selection: selection in choosing X
IDENTIFY CAUSAL EFFECTS VIA:
RCT; randomized control trails
= You compare the outcome of an individual who received a treatment X to the outcomes of
the same individual who not received said treatment. Bv. what does access to childcare do
to your income?
- Selection can occur because individuals who received ‘treatment X’ may differ from
that wo do not
o è you can eliminate this by randomly assigning individuals (who shouldn’t
differ in their observable characteristics like gender, age, ..) to treatment and
control groups and compare these two. (balanced sample)
§ The bigger a sample, the greater the probability that on average, the two
sub samples are balanced.
- Law of Large Numbers
o as sample size increases, the sample average of a given variable approximates
the population average.
§ Helps with balance even in unobservable characteristics.
! Limitations of RCT’s:
- Limited external validity: individuals behave different in experiments so you can’t
generalize the effects because they know they have been observed.
- Lacking of internal validity: spillovers to people in the control group (Dit betekent dat
de effecten die je in het experiment ziet ook te maken kunnen hebben met andere
factoren dan alleen wat je onderzoekt), selective uitval (mensen kunnen uit het
onderzoek stappen na aantal jaren)
- Ethical concerns: giving someone ‘no treatment’ can be illegal (bv. health insurance)
- Very expensive: impossible to pay for everyone’s treatment
Natural experiments
= naturally occurring variation in treatments to estimate causal effects (no impact of
researchers) bv. effects of change in law/effect of a natural disaster on society
, regression
! Challenges of natural experiments
- No clean randomization of individuals
- No balanced samples in reality
ariate regression expresses the dependent variable Yi as a linear combinat
è leads us to econometric methods to estimate causal effects: regression analysis
ercept ↵, the explanatory variable Xi and an error term ui :
Review: How does regression analysis work?
UNIVARIATE REGRESSION Yi = ↵ + ⇤ Xi + ui
Basic principle: the relation between two or more variables as a linear equation
ms
sses of
the the potential
dependent variable outcomes framework:
Yi as a linear combination of
Dependent
y variable Xi and an errorvariable:
term ui :Y (what we like to explain)
ntercept: ↵ Independent/explanatory
= E(Y ), e.g. expected
Error term:0,ui
health score in the absence of health insurance.
variable: X (values that shapes the dependent variable)
Yi = ↵ +
reatment ⇤ Xi + ui = (Y
effect: Y0,i ), e.g. difference in health scores with and without ins
1,i
= treatment effect
omes framework:
Statistical inference: How likely is it that our estimates reflect a true correlation? How
ation: precise
pected health score in are
the our estimates?
absence of health insurance.
Y ), e.g. difference in health scores with and without insurance.
est
0,i
estimator ˆ: average
Measures
↵ health
of statistical score of the control group.
precision:
- Standard error: attached to each parameter estimate (bv. standard deviation)
est estimator- ˆ:Confidence
averageinterval:
difference between
the interval thetrue
in which the scores offalls
parameter treatment and control gr
with % probability
- T-statistic:
ealth score of the control group. ratio of the parameter estimate to its standard error used in hypothesis
tests
fference between the scores of treatment and control group.
- P-value: the probability of obtaining a t-statistic at least as large as the observed one,
if the null hypothesis is correct
Statistical significance: we say an estimate is statistically significant from zero if the t-
statistic is greater than 2/ the p-value is less than 0,05
�� / ��
à the larger the sample, the greater the likelihood of precise and significant results
MULTIVARIATE REGRESSION
= If we have more than one explanatory variable X, Z, … These are variables that are potential
influencing factors on Y. We can control for the observable omitted variables by including
them in the regression.
Omitted variables
= influencing factors are correlated with both the explanatory and the dependent variable.
They bais our estimates and lead to faulty conclusions
è we cannot control for unobserved omitted variables, such as intelligence or motivation so
here we can use natural experiments again
Les 1: Introduction to labour economics
What is labour economics?
Why do we care about labour markets?
- Work makes an important part of each individuals life
o Income generating, that determines consumption
o Together with education, health, …
§ If you don’t work you don’t have money, you can’t consume anything,
you have poor health and education
- Labour is key for production
o Influences innovation and economic growth
- The labour market is subject of many social policies
o bv. social security, immigration, …
Who are the actors of the labour market?
Firms
= labour demand; how many workers to hire, wages, prices to set, invest in technology?
o maximise profits
o constraints
§ labour and capital = limited inputs
§ competition and consumer demand holds back demand
Workers
= labour supply; how many hours to work, effort at work, skills, ..
o maximise their well-being
§ utility from leisure, consumption, health, …
o constraints
§ hours in a day = limited
§ income and job offers
Government
= rules of the labour market
o maximise social welfare
o constraints
§ tax income
§ elections
o their own policy tools
§ minimum wage, health and safety regulations, immigration, taxation
via income, payroll
, 4 Chapter 1
FIGURE 1-1 Supply and Demand in the Engineering Labor Ma
The labor supply curve gives the number of persons willing to supply th
wage. The labor demand curve gives the number of engineers that firms
where supply equals demand, so that 20,000 engineers are hired at a wa
Labour supply curve: upwards sloping, Earnings ($)
workers are willing to work more if they Labor Supply
receive higher wages. Curve
50,000
Labour demand curve: downwards sloping,
firms are willing to hire more workers if they
can offer lower wages. Equilibrium
40,000
è No government intervention: the
intersection of the supply and demand Labor Demand
Curve
curves (only) determines market clearing 30,000
wages and equilibrium.
Employment
10,000 20,000 30,000
How do we study labour markets? when labor is expensive. The relation between
DATASOURCES firms are willing to hire is summarized by th
in Figure 1-1. As drawn, the labor demand c
industry want to hire 20,000 engineers when th
- Cross-sectional household surveys
engineers if the wage rises to $50,000.
o Current population survey (US)
Workers and firms, therefore, enter the lab
o European labor force survey workers are willing to supply their services
o Census willing to hire them. Conversely, few worker
the wage is low, but many firms are looking f
- Longitudinal household surveys firms search for workers, these conflicting des
o Panel study of income dynamics reaches an equilibrium. In a free-market eco
o British cohort study equals demand.
o German socio-economic panel As drawn in Figure 1-1, the equilibrium wa
hired in the labor market. This wage–employm
- Administrative data it balances out the conflicting desires of wor
o Taxation data the engineering wage was $50,000—above e
o Social registry data only 10,000 engineers, even though 30,000 e
number of job applicants would bid down th
available. Suppose, instead, that the wage w
1) Labour economic theory engineers are cheap, firms want to hire 30,00
willing to
= Simplification of the real world designed to isolate the core mechanism work at
behind onethat wage. As firms compe
specific
question. up the wage.
There is one last major player in the labor
can tax the worker’s earnings, subsidize the tra
We will make simplifying assumptions, objective function and their constraints, optimal
behavior, comparative statics (‘all else equal’)
Example: slide 14
bor04724_ch01_001-018.indd 4
, 2) Labour economic empirics
Descriptive analysis
= establish facts about economic patterns we observe in the real world, provide basis for
theory
Causal analysis
= test theoretical predictions with real world data and establish causal relationships. You
need to look for the effect of something happening with all other things equal.
! Problems with examine an effective causal relationship within the observational data:
- Endogeneity: X reacts on Y
- Omitted variable bias: Z causes effect on Y and not X
- Selection: selection in choosing X
IDENTIFY CAUSAL EFFECTS VIA:
RCT; randomized control trails
= You compare the outcome of an individual who received a treatment X to the outcomes of
the same individual who not received said treatment. Bv. what does access to childcare do
to your income?
- Selection can occur because individuals who received ‘treatment X’ may differ from
that wo do not
o è you can eliminate this by randomly assigning individuals (who shouldn’t
differ in their observable characteristics like gender, age, ..) to treatment and
control groups and compare these two. (balanced sample)
§ The bigger a sample, the greater the probability that on average, the two
sub samples are balanced.
- Law of Large Numbers
o as sample size increases, the sample average of a given variable approximates
the population average.
§ Helps with balance even in unobservable characteristics.
! Limitations of RCT’s:
- Limited external validity: individuals behave different in experiments so you can’t
generalize the effects because they know they have been observed.
- Lacking of internal validity: spillovers to people in the control group (Dit betekent dat
de effecten die je in het experiment ziet ook te maken kunnen hebben met andere
factoren dan alleen wat je onderzoekt), selective uitval (mensen kunnen uit het
onderzoek stappen na aantal jaren)
- Ethical concerns: giving someone ‘no treatment’ can be illegal (bv. health insurance)
- Very expensive: impossible to pay for everyone’s treatment
Natural experiments
= naturally occurring variation in treatments to estimate causal effects (no impact of
researchers) bv. effects of change in law/effect of a natural disaster on society
, regression
! Challenges of natural experiments
- No clean randomization of individuals
- No balanced samples in reality
ariate regression expresses the dependent variable Yi as a linear combinat
è leads us to econometric methods to estimate causal effects: regression analysis
ercept ↵, the explanatory variable Xi and an error term ui :
Review: How does regression analysis work?
UNIVARIATE REGRESSION Yi = ↵ + ⇤ Xi + ui
Basic principle: the relation between two or more variables as a linear equation
ms
sses of
the the potential
dependent variable outcomes framework:
Yi as a linear combination of
Dependent
y variable Xi and an errorvariable:
term ui :Y (what we like to explain)
ntercept: ↵ Independent/explanatory
= E(Y ), e.g. expected
Error term:0,ui
health score in the absence of health insurance.
variable: X (values that shapes the dependent variable)
Yi = ↵ +
reatment ⇤ Xi + ui = (Y
effect: Y0,i ), e.g. difference in health scores with and without ins
1,i
= treatment effect
omes framework:
Statistical inference: How likely is it that our estimates reflect a true correlation? How
ation: precise
pected health score in are
the our estimates?
absence of health insurance.
Y ), e.g. difference in health scores with and without insurance.
est
0,i
estimator ˆ: average
Measures
↵ health
of statistical score of the control group.
precision:
- Standard error: attached to each parameter estimate (bv. standard deviation)
est estimator- ˆ:Confidence
averageinterval:
difference between
the interval thetrue
in which the scores offalls
parameter treatment and control gr
with % probability
- T-statistic:
ealth score of the control group. ratio of the parameter estimate to its standard error used in hypothesis
tests
fference between the scores of treatment and control group.
- P-value: the probability of obtaining a t-statistic at least as large as the observed one,
if the null hypothesis is correct
Statistical significance: we say an estimate is statistically significant from zero if the t-
statistic is greater than 2/ the p-value is less than 0,05
�� / ��
à the larger the sample, the greater the likelihood of precise and significant results
MULTIVARIATE REGRESSION
= If we have more than one explanatory variable X, Z, … These are variables that are potential
influencing factors on Y. We can control for the observable omitted variables by including
them in the regression.
Omitted variables
= influencing factors are correlated with both the explanatory and the dependent variable.
They bais our estimates and lead to faulty conclusions
è we cannot control for unobserved omitted variables, such as intelligence or motivation so
here we can use natural experiments again
