Labor Markets in Practice: Theory, Empirical Evidence

and Policies (56891)

Lecture Note 4
Wage Differentials—Why Are People Paid Differently?

Moshe Buchinsky

Department of Economics, Sciences Po

Spring, 2025
Introduction

Wage distribution in France in 2017. Wages are net of social contributions

Note: Full time workers are workers whose labor contract stipulates 35 to 39 hours of work per week.
Introduction

Very important wage differences −→ Why are there wage

differentials?
Between firm competition on labor market
=⇒ Productivity
=⇒ Working conditions
Limited competition on labor market
=⇒ Exploitation of workers by firms
=⇒ Discrimination
=⇒ Trade unions
This lecture is focused on the effects of between firm competition
Outline

Wage differentials and productivity

Wage differentials and working conditions
Conclusion
Wage differentials and productivity

Basic principle
Between firm competition =⇒ workers are paid at their marginal
productivity
Example: A worker can produce 20 euros per hour, firms compete to
attract the worker =⇒ the worker gets 20 euros
This suggests that wages depend on productivity
Wage differentials and productivity

Consequences of the relation between wage and productivity

More educated people are more productive
=⇒ positive relation between years of education and wage
More experienced people are more productive
=⇒ positive relation between experience and wage
Wage differentials and productivity
Education

Wage distribution in France in 2017

Note: wages are net of social contributions.

Source: Enquête emploi 2017, INSEE.
Wage differentials and productivity
Education

The relation between education and wage can be represented by the

following equation
wi = a1 + a2 si + εi , (1)
i is for the ith individual;
wi : wage;
si years of education (‘s’ for school);
εi is a residual term
Wage differentials and productivity
Education

The coefficients aj , j = 1, 2, are estimated by minimizing the sum of

square residuals: ∑i εi2 .
We have many observations in the dataset: i = 1, ...N.
Minimize
N
∑ (wi − a1 − a2 si )2 .
i=1

That is, choose the a1 and a2 the minimize the sum of squared
residuals.
This is the Ordinary Least Square (OLS) method
Wage differentials and productivity
Education

If we estimate equation Wage Differentials and Productivity with

OLS, we get the prediction of wi

w
bi = 1037.948 + 55.64si , R 2 = 0.07
(18.13) (1.30)

Coefficient of determination—R 2 : Share of the variance of the

explained variable wi explained by the explanatory variable si
The variance measures the dispersion of the variable:

1 N
= ∑ (wi − w )2 ,
N i=1

where w is the mean of all wi , i = 1, .., N

Wage differentials and productivity
Education

w
bi = 1037.948 + 55.64si , R 2 = 0.07
(18.13) (1.30)

The numbers in parentheses below the coefficients are the standard

errors of the coefficients because the value of each coefficient is not
known with certainty
The standard error give idea about how close to the truth (or how
accurate) is the estimate of the parameters.
The standard error are used the to determine confidence intervals
A 95% confidence interval for a2 :

[55.64 − 1.96 × 1.30, 55.64 + 1.96 × 1.30] = [53.1, 57.2]

In other words, with 95% confidence the parameter a2 belongs to

[53.1, 57.2]
Wage differentials and productivity
Education

The residuals from the estimation:

εi = wi − w
b bi .

How spread are they?

Wage differentials and productivity
Education

This figure shows the residuals (i.e., the εi ) of the OLS estimation of
equation (Wage Differentials and Productivity).
Wage differentials and productivity
Labor market experience

Source : Enquête emploi 2017, INSEE

Wage and years of labor market experience in France in 2017. Wages are net of
social contributions
Wage differentials and productivity

Mincer’s equation:∗
Assume that wages depend on education and experience according to
the following relation:

ln(wi ) = a1 + a2 si + a3 xi + εi ,

where i is for the ith individual i;

wi = wage of individual i;
si = years of education of individual i;
xi = years of labor market experience of individual i;
εi is a residual term

∗ Source: Mincer, J. 1974. Schooling, experience and earnings, New York, NY: National Bureau of Economic Research
Wage differentials and productivity

ln(wi ) is the natural logarithm =⇒ dwi / wi = a2 dsi ,

or 1 more year of education =⇒ a2 percent increase in wage
Alternatively, the return to education
Similarly a3 is the return to experience
Wage differentials and productivity

Actually Mincer analyzed a more complete relation to account for

potential non-linearity:

ln(wi ) = b1 + b2 si + b3 si2 + b2 xi + b3 xi2 + εi , (2)

or for potential non-linearity and interaction:

ln(wi ) = b1 + b2 si + b3 si2 + b4 xi + b5 xi2 + b6 si xi + εi . (3)

Wage differentials and productivity
The Mincer equation for France in 2017

(1) (2) (3)

ln(w ) ln(w ) ln(w )

s 0.0277*** 0.0439*** 0.0811***

(0.000662) (0.000699) (0.00457)
s2 -0.00129***
(0.000151)
x 0.00893*** 0.0183***
(0.000170) (0.000631)
x2 -0.000200***
(1.35e-05)
Constant 7.066*** 6.638*** 6.300***
(0.00925) (0.0120) (0.0345)

Observations 24,723 24,723 24,723

R2 0.066 0.160 0.172

Standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1
Wage differentials and productivity
The Mincer equation for France in 2017

In the previous table, we see : ∗∗∗ p < 0.01, ,∗∗ p < 0.05, ,∗ p < 0.1.
What are these?
These are the p-values for the null hypothesis that the related
coefficient is equal to zero
The p-value is the probability to be wrong if one rejects the null
hypothesis
p < 0.01 means that the p-value is smaller that 0.01, i.e., that
probability to be wrong if one reject the hypothesis that the
coefficient is equal to zero is smaller than 1
Wage differentials and productivity

The seminal Mincer equation

Source: Mincer, J. 1974. Schooling, experience and earnings, New York, NY: National Bureau of Economic Research.
Wage differentials and productivity

The seminal Mincer equation. In the table we see that:

Below the coefficient t-statistics in parenthesis, equal to the value of
the coefficient divided by its standard error denoted by σ
If the t-statistics is larger than 1.96, then the p-value for the null
hypothesis that the related coefficient is equal to zero is smaller than
5%.
This is because the 95% confidence interval of a coefficient a, whose
estimate is b
a is given by

a − 1.96 · σ ,b
[b a + 1.96 · σ ] .

If number of observations > 100:

I p-value < 0.01 if t-stat> 2.33;
I p-value < 0.05 if t-stat > 1.96;
I p-value < 0.1 if t-stat > 1.65 (see Table)
Wage differentials and productivity

The seminal Mincer equation

If the t-statistics is at least equal to 1.96 in absolute value, the
coefficient is different from zero with confidence level of at most 5%,
“meaning” that the probability to be wrong when one claims that the
coefficient is different from zero is smaller than 5%.
If the t-statistics is at least equal to 2.33 in absolute value, one says
that the coefficient is different from zero at the 1% confidence level.
Wage differentials and productivity

Consequence of the relation between wage and productivity

I Empirically, market forces, i.e., between firm competition
=⇒ more productive workers are paid more
I This leads to:
F positive relation between years of education and wage
F positive relation between experience and wage
May imply very important wage differentials: CEO, football players ...
Wage differentials and productivity
Superstar economics

Median compensation of CEOs and other top executives, from 1936 to 2005

Note: The CEO is identified as the president of the company in firms where the CEO title is not used. “Other top executives”
includes any executives among the three highest paid who are not the CEO. All dollar values are in inflation-adjusted 2000
dollars.
Source: Frydman, C., & Saks, R. J. 2010. Executive compensation: A new view from a long-term perspective, 1936–2005.
Review of Financial Studies, 23(5), 2099-2138.
Wage differentials and productivity
Superstar economics

The variation in CEO pay is found to be mostly due to variation in

firm characteristics, whereas implied differences in managerial ability
are small and make relatively little difference to shareholder value∗
Small differences in ability can induce very large remuneration
differentials
CEO remunerations do not depend on firm characteristics only
I Potential rents due to information asymmetries
I Regulations which may affect the ability to capture the value added to
the firm by CEO

∗ Source:Edmans A, Gabaix X, Jenter D, 2017. Executive Compensation: A Survey of Theory and Evidence In: Handbook of
the Economics of Corporate Governance, Chapter 9. ; 2017. pp. 383-539.
Wage differentials and working conditions
According to Adam Smith:∗∗

Compensating wage differentials arise to compensate workers for the

non-wage characteristics of jobs

Compensating wage differentials theory: A job is a package including

the wage, but other pecuniary and non-pecuniary elements.
These elements are (among others):
I Working conditions : worse working conditions =⇒ higher wages
I Job protection: more job stability =⇒ lower wages
I Health insurance: more health insurance =⇒ lower wages

∗∗ Adam Smith, 1776, An Inquiry into the Nature and Causes of the Wealth of Nations.
Wage differentials and working conditions
Application of compensating wage differentials theory:

More risky jobs =⇒ Higher wages

Effects of the AIDS epidemic on the wages of nurses in the United

States
I Nurses practicing their profession in zones where the epidemic is
especially prevalent have greater risk of contracting the disease than
nurses in zones where AIDS is less prevalent
I Comparison of the earnings of nurses, the test group, with those of
college-educated non-health-care workers, the corresponding control
group
The finding is that 10% increase in the prevalence of AIDS
=⇒ increase in nurses’ wages of slightly less than 1% relative to the
wages of the control group
Wage differentials and working conditions

The relation between wage differentials and working conditions reveals

information about the preference of individuals for working conditions
This provides a way to evaluate the price of human life!
Suppose each person in a sample of 100,000 people were asked how
much she/he would be willing to pay for a reduction in their
individual risk of dying of 1 in 100,000, or 0.001%, over the next year.
Since this reduction in risk would mean that we would expect one
fewer death among the sample of 100,000 people over the next year
on average, this is sometimes described as "one statistical life saved.”
Wage differentials and working conditions

Now suppose that the average response to this hypothetical question

was $100.
Then the total dollar amount that the group would be willing to pay
to save one statistical life in a year would be $100 per person times
100,000 people, or $10$ millions.
This is what is meant by the "value of a statistical life.”∗∗

∗∗ Source: United States Environmental Protection Agency.

Wage differentials and working conditions

Source: Thaler, R., & Rosen, S. (1976). The value of saving a life: Evidence from the labor market, In N. Terleckyj (Ed.),
Household production and consumption. New York, NY: Columbia University Press.
Wage differentials and working conditions
Value of Statistical Life

Evaluation of the value of statistical life

Source: OECD, 2012. The value of statistical life: a meta-analysis, Working Party on National Environmental Policies
Conclusion

Large wage differentials, influenced by supply and demand

=⇒ productivity is a very important factor ... but not only
Working conditions play also an important role ... but also
I Market power of firms
I Discrimination
I Trade unions
READING:
George Borjas. 2013, Labor Economics, 6th edition, pages 12-20 on
regression analysis, and Sections 5.3 to 5.6 in Chapter 5.
End of Lecture note 4