Returns to Talent and the Finance Wage Premium

Returns to Talent and the Finance Wage
Premium∗
Claire Célérier
†
Boris Vallée
‡
First Draft: October,14 2010
This Draft: September 10, 2015
Abstract
We study the role of talent in the distribution of pay in the finance industry since the
1980s. We exploit a special feature of the French educational system to build a precise
measure of talent that we match with compensation data for graduates of elite French
institutions. Wage returns to talent are three times higher in the finance industry than in
the rest of the economy. This greater sensitivity to talent almost fully absorbs the finance
wage premium, as well as its increase, since the 1980s. Finally, the share of variable
compensation is positively correlated with returns to talent.
Keywords: Finance, Compensation, Talent, Wage Distribution, Wage Structure, Superstars
JEL codes: G2, G24, J3, J31, M5
∗
We gratefully acknowledge helpful comments and suggestions from Philippe Askenazy, Bruno Biais, Gilbert Cette,
Vicente Cunat, Rudiger Fahlenbrach, Thierry Foucault, Francois Geerolf, Mireia Gine (EFA discussant), Vincent Glode,
Ulrich Hege, Johan Hombert, Steve Kaplan (NBER discussant), Augustin Landier, Arnaud Maurel, Steven Ongena, Evren
Ors, Thomas Philippon, Thomas Piketty, Josh Rauh (AEA discussant), Jean-Charles Rochet, David Scharfstein, John
Thanassoulis, David Thesmar, Alexander Wagner, Luigi Zingales and seminar and conference participants at Stanford
University, Toulouse School of Economics, Paris School of Economics, University of Mannheim, Banque de France, University
of Zurich, Rotman School of Management, Solvay Brussels School, Lugano Institute of Finance, the 2015 NBER Corporate
Finance meeting, the AEA 2015 annual meeting, the European Labor Association conference, the EFA 2014 annual meeting,
the AFFI annual meeting, the URPP Finreg conference, and the CSEF. We are very grateful to IESF and Chantal Darsch
for providing data, and we also thank her for all her help and support.
†
Claire Célérier - University of Zurich, E-mail: [email protected] (Contact Author).
Claire Célérier acknowledges financial support from the Swiss Finance Institute and URPP-Finreg.
‡
Boris Vallée - Harvard Business School, Email: [email protected].
1
1
Introduction
Compensation in the finance industry has been higher and more skewed than in other
sectors since the beginning of the 1980s. Controlling for education and other individual
characteristics, Philippon and Reshef (2012) find that the finance wage premium is, on average, 50% for 2006. The financial sector has therefore largely contributed to the observed
gains at the top of the wage distribution since the 1980s (Kaplan and Rauh (2010), Bell
and Van Reenen (2013)) and, consequently, has often been criticized as a source of growing income inequality. This public debate, associated with increased regulatory scrutiny,
calls for an improved understanding of the drivers of bankers’ pay. A growing theoretical
literature relies on talent as a key ingredient in models of bankers’ pay (Acharya et al.
(2013), Benabou and Tirole (2015), Glode and Lowery (2015), Thanassoulis (2012)).
However, empirically assessing the returns to talent across industries is difficult because it requires accurately observing and measuring worker talent. A unique feature of
the French educational system is that prospective engineering students are selected solely
based on their performance on a nationwide competitive exam that covers a wide range
of subjects in both written and oral formats. We exploit this rigorous, multi-dimensional
selection process to build a measure of talent, which we use to address our research question: Are wage returns to talent relatively high in finance compared to the rest of the
economy?1 More broadly, do talent effects drive the cross-section of wages observed in
finance?
Using our novel measure of talent, we show that the returns to talent are three times
higher in the finance industry than in the rest of the economy, and that they almost fully
absorb the finance wage premium once included in the regression specifications. Returns
1
For the purposes of this analysis, we define talent as the aptitude to reach an objective in a competitive environment. Talent hence encompasses not only cognitive skills but also non-cognitive skills
and personality traits, such as motivation, self-discipline, low cost of effort, and ability to perform in a
competitive environment.
2
to talent have also increased threefold since the 1980s, which is in line with significant
growth in bankers’ pay over this period. Additionally, we show that talented workers in
finance receive a relatively large share of variable compensation. We discuss our findings
in light of two complementary theories: competition for talent and moral hazard.
We use the selectivity of French engineering schools to measure the talent of its alumni
for the following reasons. The national competitive exam for engineering schools incorporates both written and oral sections covering a wide range of subjects. This exam assesses
academic, cognitive, and communication skills, and it gauges personality traits such as endurance, commitment, ambition, and ability to perform in a competitive environment.2,3,
4
The highly selective and competitive environment of preparatory schools prior to the
examination, as well as the high stakes of the exam outcome, indeed test candidate motivation and resistance to stress. In addition, with more than 35 hours of classes, heavy
homework loads, and one written and two oral exams per week, the intense workload
during these school years ensures that performance is unbiased by personal coaches, exam
preparation boot camps, and other support resources that are often used by applicants
to US universities. A further element of the suitability of our research set-up is its focus,
by virtue of analyzing talent heterogeneity in a highly educated cohort, on the right tail
of the population distribution.5 Finally, there are 225 small-scale engineering schools in
France, which provides a high level of data granularity.
We complement this school-level measure of talent, and control for school treatment
2
Ors et al. (2013) exploit this specificity of the French educational system for business schools. They
find that males’ performance dominates that of females in competitive environments, whereas females
outperform men on less competitive exams.
3
French engineering schools confer a generalist degree. The proclaimed objective of the selection
process is to identify the most talented individuals in France, and therefore, it goes beyond assessing
technical skills.
4
The literature shows that personality traits are important determinants of wages (Heckman (1995),
Bowles et al. (2001), Heckman and Kautz (2012)).
5
The heterogeneity in talent at the right tail is typically overlooked in discreet population-wide measures, such as SAT scores, while the objective of the engineering school exam is to precisely discriminate
among individuals in a highly educated and homogeneous population.
3
effects, by also considering the number of years it took a candidate to be admitted into a
given “Grande Ecole”.6 A student accepted at a top school after only one year of exam
training is likely more talented than a student who requires three years of training.
We match these talent measures to detailed compensation survey data that cover 7%
of the total population of French graduate engineers. The survey, which gathers alumni
data from 199 of 225 French engineering schools, includes detailed information on education, occupation, family situation, industry, firm type and size, and compensation. It
also provides us with the specific job title of the worker. Because engineering, business
administration and medicine are the only fields that are selective in France, and engineering is the largest of the three, this dataset covers a significant share of the right tail of
the skill distribution in the French population. Our dataset spans the period from 1983
to 2011. Each of the 15 repeated cross-sections covers, on average, 30,800 individuals
working in France or abroad. Using these survey data, we show that French graduate
engineers in the finance sector are better paid, earning a premium of 25% over our sample
period. This premium has increased threefold since the 1980s. This finding is consistent
with Philippon and Reshef (2012). In line with Bell and Van Reenen (2014) and Bell and
Van Reenen (2013), we also observe relatively high and increasing skewness in the wage
distribution in the finance industry.
The central result of our paper is that returns to talent are significantly higher in the
finance industry and that they almost entirely absorb the finance wage premium when
included in the econometric specification. The main equation in our empirical analysis
regresses the log of yearly gross wages on our talent measure and its interaction with
industry dummies. Graduating from a school one notch higher in terms of selectivity induces a 6.5% average wage premium in the finance industry versus a 2% relative premium
6
Age at graduation maps with age at entry. While a large number of students repeat their last year of
preparatory classes to improve their ranking, virtually no students skip or repeat a year during engineering
school.
4
in the rest of the economy. When we include the interactions between our talent measure
and sector fixed effects, we observe that the premium for working in the finance industry
decreases from 25% to 2.4% and is no longer significant. The finance wage premium is
thus disproportionately allocated to the most talented individuals in this industry. Within
finance, returns to talent are even higher for front office jobs than for back office or support
departments.
The foregoing result is confirmed when graduation age is used as an alternative measure
of talent, thereby allowing all unobserved school-level variables to be absorbed by school
fixed effects. We again find that wage returns to talent are three times higher in the
finance industry than in the rest of the economy and account for a significant part of
the finance premium. This additional analysis eliminates alumni networks and school
differences in quality of training or focus on finance as possible explanations for our main
result.
Our result is also robust to the introduction of individual fixed effects in a panel regression that estimates the effect on wages of switching into the finance industry from
another sector. We track individuals across surveys via detailed sociodemographic variables, such as their father’s and mother’s occupations and year of birth, and educational
variables, such as the name of their engineering school and type of specialization. We
find that the wage premium obtained from switching into the finance industry is fully absorbed when we control for heterogeneous returns to talent across industries. Therefore,
our main result should not be driven by unobserved time-invariant characteristics at the
individual level, such as social background or risk aversion. All three results are robust
to the introduction of industry-year fixed effects, which absorb any overall shift of wages
in finance.
We also observe a trend toward increasing returns to talent. Estimating our main
5
equation over each of the three decades reveals that wage returns to talent in finance
have increased nearly threefold over the 1980–2011 period. The most talented individuals
in finance have thus received most of the increase in the wage premium over the past
decades. Our results shed new light on the wage growth in finance observed since the
1980s, and documented in the literature. Increasing returns to talent are also associated
with an increasing share of talented individuals going into finance.
Finally, we show that the share of variable compensation is positively correlated with
returns to talent.7 Our findings thus point to an interaction between the returns to talent
and the structure of compensation.
We find alternative channels for higher returns to talent in finance difficult to reconcile
with the empirical facts. A battery of specific tests precludes network effects, social background factors and compensating wage differentials as potential drivers of our results. For
example, we find that returns to our talent measure are even higher for first-generation
graduates, the children of parents without university degrees. Favorable social backgrounds or personal relations are unlikely to play an important role for this subsample of
graduates.
We then discuss the possible micro-foundations for the significantly higher and increasing returns to talent we observe in the finance industry. This gap might exist due to
the heterogeneous preferences that segment the job market. The higher returns to talent
we observe in finance may thus result from intense labor market competition coupled
with a high talent scalability in finance. A growing theoretical literature hence investigates the effects of competition for talent in the finance industry (Acharya et al. (2013),
Benabou and Tirole (2015), Glode and Lowery (2015), Thanassoulis (2012) and Bijlsma
et al. (2012)).8 An alternative explanation posits that higher returns to talent result from
7
We calculate the variable wage from a survey question on compensation structure.
Financial workers may be more talented at extracting rents from society rather than being more
socially productive, as in Biais and Landier (2015) and Bolton et al. (2013).
8
6
rent extraction by finance workers from their employers. Talented workers may be more
costly to incentivize due to superior outside options or be more costly to monitor, as they
perform more opaque tasks (Axelson and Bond (2015), Biais et al. (2015)). These two
mechanisms, competition for talent and moral hazard, are not mutually exclusive and are
likely to interact (Benabou and Tirole (2015)).
Our work expands on the recent empirical literature that has identified a high level
of compensation in the finance industry relative to the rest of the economy and high
skewness at the top of the wage distribution. Philippon and Reshef (2012), Oyer (2008),
and Goldin and Katz (2008) – based on data from the Census Population Survey, a
Stanford MBA survey, and a Harvard alumni compensation survey, respectively – find
that the finance premium varies from 40% (in Philippon and Reshef (2012)) to more
than 100% (in Oyer (2008), and Goldin and Katz (2008)). Philippon and Reshef (2012)
documents a post-1980s increase in compensation in finance relative to the rest of the
private sector after controlling for education, and Kaplan and Rauh (2010) and Bell and
Van Reenen (2014) show that the financial sector share at the top end of the income
distribution has significantly increased. The main contribution of the present paper is to
attribute these wage distribution patterns in the finance industry to higher and increasing
returns to talent.
Our paper contributes to the literature that investigates the dramatic growth in top
executive pay and earning inequalities observed since the 1980s. This literature includes
theories of managerial power (Bebchuk and Fried (2004)), social norms (Piketty and Saez
(2006); Levy and Temin (2007)), incentives, and competition for talent or managerial
skills (Frydman (2007), Murphy and Zábojnı́k (2004), Gao et al. (2015), Geerolf (2014),
Guadalupe (2007)). Our results are consistent with the evolution of wages reflecting a
change in market returns to talent, magnified in recent decades by scale effects (Gabaix
7
and Landier (2008), Kaplan and Rauh (2013), and Greenwood and Scharfstein (2013)),
skill-biased technological change (Katz and Murphy (1992); Garicano and Rossi-Hansberg
(2006)), and deregulation (Boustanifar et al. (2015)).
Our paper also provides new evidence on the interaction between competition for talent
and the structure of compensation. Lemieux et al. (2009) show that wages are more closely
related to worker production in performance-pay than in non-performance-pay jobs, and
Cuñat and Guadalupe (2005) show that a higher level of product market competition
increases the performance pay sensitivity of compensation schemes. Reliance on incentive
pay may be higher for talented workers because of higher monitoring costs (Biais and
Landier (2015)), higher productivity of effort, or better outside options (Giannetti and
Metzger (2013)), but causality may also occur in the opposite direction: performance pay
may be used as a sorting mechanism to attract talented workers (Benabou and Tirole
(2015)).
Finally, the results reported in this paper on the increasing returns to talent and talent
allocation effects raise questions concerning the externalities that might be generated by
high returns to talent in the finance industry. By offering relatively high wages for the same
level of talent, the finance sector may lure talented individuals away from other industries
(which Baumol (1990) and Murphy et al. (1991) argue may have a downward impact on
economic growth) or from financial regulation (Shive and Forster (2014), Bond and Glode
(2014)). Shu (2013), however, shows that the financial industry’s talent-capture effects are
limited, and Böhm et al. (2015) find no evidence that the selection of talent into finance
has increased or improved. Competition for talent can also generate inefficient risk taking
(Acharya et al. (2013)), lead to excessive overbids (Glode and Lowery (2015)), increase
the fragility of banks (Thanassoulis (2012)), or shift effort away from less contractible
tasks, resulting in efficiency losses (Benabou and Tirole (2015)).
8
The paper proceeds as follows. In Section 2, we describe how we measure talent. In
Section 3, we provide summary statistics for our dataset and assess the representativeness
of the sample. We present our results in Section 4, and investigate the relation between
returns to talent and the structure of pay in Section 5. Section 6 rules out alternative
explanations for our results on the returns to talent in finance. Section 7 discusses and
interprets our results in light of the predictions of related theories. Section 8 concludes.
2
Measuring Talent
Measuring relative returns to talent in an industry is challenging because of how difficult
it is to accurately observe talent. Disentangling talent from confounding factors, such as
social backgrounds, especially at the top of the talent distribution, is also difficult. Endogenous industry-worker matching represents a final challenge for measuring the returns
to talent.
The ideal experiment for our study would therefore cover a highly and homogeneously
educated population wherein talent and pay are perfectly observable to the econometrician. In addition, workers should be randomly assigned to a given industry.
We argue in this section that the French educational system provides us with the closest
feasible setup to the ideal experiment, as it differs from this ideal setup along only two
dimensions and in a moderate manner. First, although talent is not perfectly observable,
the selection process of the French education system allows us to build a comprehensive
and robust proxy for talent within a highly and homogeneously educated subpopulation.
Second, although allocation to a sector is not random, each sector is represented for each
level of talent, which allows us to build robust counterfactuals.9
9
See Table A7 in the online appendix.
9
2.1
French Engineering Schools’ Selection Process
We use the selection process of French engineering schools to build a measure of talent
for the entire population of engineers. To earn the official title “graduate engineer”,
students in France need to graduate from a master’s program in any field of engineering
offered by one of 225 selective, small-scale institutions.10 These so-called “Grandes Ecoles
d’Ingénieurs” select students on the basis of their ranking on a national competitive exam
that includes both written and oral tests. Student performance on this exam reflects
strong cognitive and academic skills as well as personality traits such as motivation, selfdiscipline, low cost of effort, and ability to work under pressure. Figure 1 summarizes the
selection process of French engineering schools.
First, written tests covering a wide range of subjects assess a large set of cognitive
and academic skills, including compulsory mathematics, physics, programming, French
literature, and foreign language sections. Candidates also select an optional topic from
among biology, chemistry, engineering, or computer science. More than 80 hours of testing
is involved over a three-week period (see Figure 1 for coefficients and exam lengths for
each topic). The candidates are then ranked nationally, and they access oral examination
with schools depending on their ranking.
The second phase of the competitive exam includes a series of 20-minute oral exams,
which assess presentation, communication, and interaction skills. Candidates solve problems in the same set of subjects and present their solutions to one or more professors in
interviews.
The process concludes with the assignment of a final national ranking that assures
applicants to engineering schools a priority position. Students favor reputation over field
expertise or location in their selection of schools, and deviations are quite rare, especially
10
Thirty thousand diplomas are awarded annually at the national level. Engineering, which is considered a generalist education, has traditionally been the main pathway to obtaining management or
leadership positions in France.
10
for top schools. Admitted students study on campus for three years before being awarded
a graduate degree.
The two years that students spend preparing for the exam at highly selective institutions (Classes Préparatoires) are a fundamental part of the process. These preparatory
schools are mainly public; they are free of tuition and provide subsidized housing. They select students on the basis of superior academic performance in high school, independently
of their social or geographic origin.11 Top high school students are highly encouraged
to apply to Classes Préparatoires. These two years of study, as well as the three years
of engineering school, are virtually free, which limits concerns about selection on family
wealth.
Success in Classes Préparatoires requires a set of personality traits that have been
proven to matter professionally, such as high motivation, self-discipline, low cost of effort,
and ability to work under pressure. First, the workload is very heavy, with more than
35 hours of classes per week, one written and two individual oral exams every week, and
a large amount of compulsory homework. Second, students have the option to switch
into the non-selective French university system at any time. Third, students are ranked
quarterly, and they are excluded after the first year if their performance is too low (Ors
et al. (2013)).
INSERT FIGURE 1
2.2
School Ranking and Talent Measure
We arrive at a talent measure by classifying engineering schools into ten selectivity categories based on the nationwide competitive exam. Group 1, which enrolls, on average,
the most talented students, includes the most selective school, while Group 10 includes
11
The average selection rate in the science and engineering fields is approximately 15% for those who
hold a scientific Baccalauréat. Source: www.data.gouv.fr.
11
the least selective schools.12
We compute a school’s selection rate by dividing the rank on the national exam of the
last admitted student by the total number of enrolled students nationwide. Information
on the rank of the marginal student and on the total number of enrolled students is public
and available for the 2002–2012 period.13 The precise methodology for this calculation is
described in the online appendix.
Selection rates for each category are reported in columns (1) and (2) of Table 2. The
highest category includes the Ecole Polytechnique, which recruits the top 1.5% of students.
The second highest category includes Mines de Paris, Ecole Centrale Paris, and Ecole des
Ponts et Chaussées. The lowest category includes mainly schools that admit students
directly after high school. The bottom part of Figure 1 plots the admission rate across
the groups of our talent measure. Table A6 in the online appendix lists the rank and the
selection rate of all schools in our sample.
Our measure of talent possesses several unique advantages over previously used measures of talent. First, it covers, with high comparability owing to consistent ranking, the
total population of French engineers since 1980, allowing us to focus on the right tail
of the distribution of talent and therefore to be more precise than when using discreet
population-wide measures, such as SAT scores. Second, the measure maps not only cognitive skills but also requisite traits for successful careers, such as motivation, resistance
to stress, and interpersonal skills. Third, in terms of social prestige, and even pay-offs
(as students from the top school are eligible for stipends), the stakes of the competitive
exam are very high and comparable to those associated with professional careers. Fourth,
the homogeneity of the population we analyze enables us to disentangle education from
talent: All engineers have the same level of education and years of schooling, and they
12
Our results still hold when we use the selection rate directly as a talent measure (see Table A4 in the
online appendix).
13
http://www.scei-concours.fr/.
12
followed the same educational path (they pursued a science major in high school and successfully applied to a selective preparatory school). Each student self selects, with respect
to personal investment, to sit for the toughest of exams, despite guaranteed admission
to a French university in any year following their high school graduation. Finally, the
admission process is designed to limit distortions due to networking, social background,
reputation, and donations, as the written exam is completely anonymous and there is no
use of letters of recommendation. The heavy workload imposed on all students during
the two years of preparation limits the potential benefits of additional resources, such as
tutors or exam preparation boot camps.
2.3
Controlling for Treatment Effects: A Non-school-specific
Measure of Talent
We also use the student’s age at graduation as an alternative measure of talent, which
enables us to differentiate among graduates within each school. In the French educational
system, high-performing students, on average, graduate at a relatively younger age both
because they skip a year and because less talented students often repeat years.14 Hence,
a student who enters the first-ranked engineering school, the Ecole Polytechnique, at the
age of 19 after two years of preparation will be more talented, on average, than a student
who enters the school at the age of 21 after three years of preparation. Age at graduation,
which is not school specific, enables us to control for unobserved school variables by
introducing school fixed effects.15 Figure A4 in the online appendix plots the distribution
of graduation age in our sample.
14
As many as 25% of students preparing for engineering schools repeat their second year of preparation
to improve their results on the competitive exam.
15
For instance, schools might offer training of varying quality or a more specific focus on finance.
13
2.4
Industry- and Job-Employee Matching
Another advantage of using the French engineering schools’ selection process to measure
the returns to talent across industries is that a small fraction of graduates of each school go
into the financial sector, which allows the building of credible counterfactuals both across
industries for the same level of talent and across levels of talent for the same industry.16
In addition, our setup is immune to concerns over self-selection into finance of the bestranked graduates within a given level of talent. By construction of the selection process,
the “bottom” graduates of a given talent level are indeed more talented, as measured by
the competitive exam, than the top graduates of the level below.
Another potential concern would be that, within a talent level, a specific subgroup,
for instance, the most social students (as suggested by Shu’s (2014) results), are going
into finance, whereas in lower talent groups, the opposite selection effect occurs, i.e., the
least social individuals are going into finance. Such a selection effect could only drive our
results over 10 groups of talent if it is itself highly correlated with our measure of talent,
which seems unlikely. Including specifications with individual fixed effects also mitigates
these concerns.
Finally, endogenous matching may also happen at the job level, as talented people are
more likely to obtain certain jobs. We address this issue by also conducting our analysis
at the job title level.
16
See Table A7 in the online appendix for the distribution of respondents across industries and talent
levels.
14
3
Data
3.1
Survey
We empirically analyze the results of a detailed wage survey consisting of 324,761 observations of engineering school graduates from 1983 to 2011. The survey, conducted by
the French Engineering and Scientist Council (IESF), a network of alumni organizations
representing 199 of the 240 French engineering schools, or 85% of the total population
of French graduate engineers in 2010, solicits the latest yearly gross wages, as well as
detailed information on the demographics, education, career, job position, and employer,
of each graduate.17,18
The survey explicitly asks for the yearly gross wages available on the latest December
pay sheet and for the employer’s five-digit industry code. Yearly gross wages therefore include cash bonuses but exclude compensation as stocks or options. Incentives to misreport
due to tax concerns are low, as in France, firms directly declare wages to the tax authority,
and the survey is anonymous. In addition, we retain only observations accompanied by a
valid industry code to ensure that respondents actually consulted their pay sheets, which
thereby maximizes the accuracy of wage data and limits measurement error. Since the
year 2000 survey, respondents have also reported the share of total compensation that is
variable, which includes cash bonuses but again excludes stocks and options.19 Cleaning
the survey data leaves us with 247,201 observations.20
17
http://www.iesf.fr/.
Source: French Education Ministry.
19
If anything, we should expect this potential measurement error to bias our results on the pay structure
downward.
20
We clean the survey data by retaining only respondents between the ages of 20 and 65 who are
full-time employees, possess a valid industry code, and have more than one year of experience. We
exclude respondents whose compensation is less than the legal minimum wage, and for each sector and
year, we drop compensation above the top 1% of the distribution. We do not trim the data at the
total sample level so that highly paid sectors are not overrepresented in the affected subsample. The
results, however, are robust to using the total sample without dropping any observation (see Table A3
in the online appendix). Finally, all nominal quantities are converted into constant 2005 euros using the
French National Price Index (IPCN) from the French National Statistical Institute (INSEE). The data
are available at http://www.imf.org/external/datamapper/index.php.
18
15
Our analysis benefits from several key features of the IESF survey. Its provision of the
name of the engineering school from which each respondent graduated is essential to the
implementation of our talent measure. Its access to unique wage data, including information on the variable share, is key to our analysis. Finally, the substantial information the
survey provides on demographics, job positions, employers, and work locations (including
engineers working outside of France, for example, in London or New York) enables our
analysis to control for a broad set of variables.
3.2
Summary Statistics
Table 1 provides key variable summary statistics together with information on the scope
of the survey. Its frequency increased from every five years from 1983 to 1986 to annually
beginning in 2004. The final number of respondents per survey averages 17,885, and each
survey represents, on average, 6.9% of the total population of French engineers. The
response rate is 18.8%. Although response is voluntary and the survey is sent only to
alumni whose names and addresses are known to the association, selection effects are
likely to be low. First, the median gross wage including bonuses reported in the 2009
survey is similar to that computed for the same population in a 2009 survey of French
companies conducted by Towers Watson, a leading compensation consulting company.
Second, respondent demographics are similar to those obtained by the INSEE in the
French Employment Survey whose sample is randomly selected. Third, selection effects
would bias our analysis if and only if respondents declare wages differently according to
both their talent level and their industry. If anything, we would expect these selection
effects to bias our results downward. Talented individuals working in industries with high
returns to talent would have indeed fewer incentives to respond to the survey because
their opportunity cost is higher.21
21
The IESF mailed the survey until 2000, and it has e-mailed the survey since 2002.
16
INSERT TABLE 1
The wage distribution among French graduate engineers has become increasingly scattered over the past three decades. Whereas the average wage, in constant euros, decreased
slightly in our sample, from 63,000 euros in the 1980s to 58,000 euros in the 2000s, due
to composition effects, the wages at the 99th percentile increased by more than 14% over
the same period.22 This result is in line with recent research showing that inequality
has increased in most OECD countries, mainly at the very top of the wage distribution
(Piketty and Saez (2003); Piketty and Saez (2006)).
We define 48 industries based on the official industry classification codes respondents
provided. Table 1 details the percentage share of respondents in the highest-paying industries (i.e., finance, oil, chemical, and consulting). Finance accounts for approximately
2% of the total sample.23
Table 1 also includes summary statistics on demographics, jobs, careers, employers,
work locations, and compensation structures. The decrease in respondents’ average age
is likely driven by the change to an e-survey format. The increase in the share of female
respondents is in line with the evolution of the composition of the engineer population nationwide. The share of respondents working outside of France has dramatically increased,
which is consistent with the improved mobility of highly qualified workers.24
3.3
The Talent Measure
Table 2 reports the selection rate (column (2)) and the number of schools and students
(columns (3) to (5)) by talent category. Our talent measure is available for 226,846
observations. By construction (of our talent measure), a larger number of respondents
22
The slight decrease is due mainly to the decrease in the age of the average respondent.
See Table A7 in the online appendix for a detailed list of, and the distribution of workers across, all
industries.
24
See the online appendix for a list of the questions asked in the 2008 survey.
23
17
is associated with a lower level of talent. Based on column (6), which reports the share
of respondents that graduated at least one year earlier than the standard age by talent
category, the age at graduation appears to be highly correlated with the level of talent.
Its focus on a highly educated population notwithstanding, our sample offers considerable
heterogeneity with respect to talent.
INSERT TABLE 2
We confirm that our talent measure is correlated with individual productivity by focusing on a specific field for which proxies for productivity are available: academic research.
We proxy output using citation counts from Google Scholar. We collect the information
for all alumni from the top 16 French engineering schools who work as academics in US
universities. We use LinkedIn and searches of top university departments to identify these
alumni. Column (7) of Table 2 provides the predicted citation count for the 16 schools
that correspond to the four highest levels of our talent measure. The predicted values
result from an OLS regression that controls for the research field, gender, experience, experience squared and experience cubed. We observe that the number of citations indeed
increases significantly with our talent measure.
3.4
Representativeness of the Sample
We implement two checks to assess potential selection effects in our survey data. First, we
benchmark our worker survey data to employer survey data from Towers Watson covering
the French engineer population. Second, we compare the patterns of compensation in the
finance industry observed in our data to those found in the literature. In both cases, we
find highly similar statistics. In addition, we find that the demographic characteristics of
respondents have evolved over time in a similar way in finance and in other sectors.
18
Figure A1 in the online appendix plots the median starting salary from both our
worker survey and the Towers Watson employer survey by industry and level of talent.25
We observe that wages are similar for both surveys along these two divisions of the data. If
anything, employers seem to report slightly higher wages than workers. As in any survey,
respondents may misreport their income; typically, this takes the form of exaggerating
income. However, this bias should be limited by the anonymous nature of this survey. In
addition, this bias is also likely to be relatively constant across sector and level of talent.
We then replicate the main stylized facts from the literature on bankers’ pay using
our data. Figure 2 plots the evolution of the coefficient of the finance sector dummy in
quantile regressions estimated at the 10th , 50th , and 90th percentiles in the 1980s, 1990s,
and 2000s samples. Skewness in wages in our sample has increased significantly over
the past decades, as already documented in Philippon and Reshef (2012) and Bell and
Van Reenen (2014).26
INSERT FIGURE 2
We confirm this observation by estimating the annual wage premia in the finance
industry via the following equation:
wi,t = × T alenti + β × Ii + γ × Xi + µ × Dt + λi,t ,
(1)
where wi,t is the log of yearly gross wages, T alent is the talent measure, Ii represents the
vector of industry dummies, Dt is the vector of year dummies, Xi is a vector of individual
characteristics, and represents the average returns to talent in the economy.27 This
25
We can only compare our data to the survey for starting salaries.
See Figure A3 in the online appendix for a description of the evolution of wages at the 10th , 50th ,
and 90th percentiles of the earnings distribution in the finance, oil, chemical, and consulting industries.
27
For the purposes of clarity, and so that it increases with worker skill, T alent is defined in our main
measure as 10 minus the rank of the school from which a respondent graduated.
26
19
estimation controls for our talent measure as well as for demographic, occupation, job,
and employer characteristics.28,29
The results are displayed in column (1) of Table 3. The average wage premium in
finance over the 1983–2011 period in our sample is 25%, compared to 14%, 13%, and 7%
in the next best paying industries, consulting, oil and chemistry, respectively. Our finding
that finance industry workers are the best paid is consistent with the results reported by
Philippon and Reshef (2012), Oyer (2008), Goldin and Katz (2008). Our estimation of the
finance wage premium is in the lower range of recent estimations reported in the literature,
which is likely due to our rich set of controls, most importantly, our talent measure, and
to the educational homogeneity of our sample. In addition, the explanatory power of
our talent measure is relatively large compared to the usual measures in the literature,
such as the principal component of the US Armed Services Vocational Aptitude Battery
(ASVAB) tests or the Armed Forces Qualifying Test (AFQT).30
INSERT TABLE 3
The external validity of our sample is further supported by Table A2 in the online
appendix, which replicates Table 6 from Bell and Van Reenen (2014). The first column
of Table A2 of the appendix shows that the finance wage premium has increased from
28
Acemoglu and Autor provide evidence of the strong explanatory power of occupational categories in
wage regressions.
29
Demographic controls include years of experience, experience squared, experience cubed, gender,
marital status, and gender×marital status. We control for occupation using nine dummies (production,
logistics, development, IT, commercialization, administration, executive, education) and for employer
type using five dummies (self-employment, private sector, state-owned company, public administration,
and other (e.g., non-governmental organizations)), and for firm size with four dummies (fewer than 20,
from 20 to 500, from 500 to 2,000, and more than 2,000, employees). Job characteristics are represented
by an “Ile de France” dummy (Paris area), a working abroad dummy (as well as country dummies for
the United States, the United Kingdom, Germany, Switzerland, Luxembourg, China, and Belgium from
2004), and four hierarchical responsibility dummies from no hierarchical responsibility to chief executive.
Table A1 in the online appendix displays the coefficient of these control variables.
30
After controlling for experience and education, we indeed find that adding our talent measure induces
an increase in the R2 four times higher than that induced by the usual talent measures (Bowles et al.
(2001)).The change in the R2 is 2.8% in our specification versus a median value of 0.7% in the literature.
This gap is likely to be underestimated because of our precise control for education. This relatively
large explanatory power is likely due to the inclusion of personality traits, which are an important and
increasing determinant of wages (Heckman (1995), Bowles et al. (2001), Heckman and Kautz (2012),
Deming (2015)).
20
7%, on average, in the 1980s, to more than 30%, on average, over the 2004–2011 period,
and the premium is much higher at the 90th than at the 10th and 50th percentiles of the
wage distribution. The last row of the table shows the average annualized increase in the
premia to be more than 2.8% at the 90th , less than 0.7% at the 50th , and 0.3% at the 10th ,
percentiles. Our finding that the finance wage premium has increased dramatically since
the 1980s and that it is concentrated among top earners is also consistent with Philippon
and Reshef (2012) and Bell and Van Reenen (2014).
4
Results
4.1
Heterogeneous Returns to Talent across Industries
We report here our central result that returns to talent are significantly higher in the
finance industry and almost entirely absorb the sector’s wage premium.
Graphical evidence of this result is provided in Figure 3, which plots respondents’
predicted wage by industry over the ten categories of our talent measure using equation
(1). We observe wages to be an increasing function of talent, and the magnitude of
this relationship is significantly higher in the finance industry than in other sectors. For
example, wages increase from the bottom to the top of the talent distribution in the
finance industry by more than 64%, while in the oil industry, they increase by only 35%.
The relationship between our talent measure and wages in finance appears to be convex.
INSERT FIGURE 3
We specifically test whether and to what extent wage elasticity to talent is high in
finance by including interactions between talent and each industry dummy in equation
(1):
wi,t = × T alenti + β × Ii + × Ii × T alenti + γ × Xi + µ × Dt + λi,t ,
21
(2)
where is the industry-specific component of returns to talent (the other variables are
the same as in equation (1)).
Column (2) of Table 3 reports the results. The positive and significant coefficient
of the interaction term between the finance dummy and talent measure shows returns
to talent to be significantly higher, three times higher in fact, in the finance industry
than in the rest of the economy. Moving one notch up our talent scale yields a 6.3%
increase in wages for a finance worker vs. 1.9% for a worker in the rest of the economy.
The consulting industry, consistent with its high talent scalability, offers returns to talent
twice as high as in the rest of the economy. Conversely, returns to talent are significantly
lower in the oil and chemical industries than in the rest of the economy, likely because of
strong physical constraints that limit the scalability of talent in those sectors.
High returns to talent in the finance industry almost entirely absorb the finance wage
premium. When we include the interaction term Ii × T alenti,t in our specification, the
finance premium almost disappears and, at 2.4%, is no longer significant (column (2)).
This result is strongly supportive of talent effects driving the finance wage premium and
is robust to including industry time year fixed effects (column (3)) and alternative talent
measures, such as 1–school selectivity rate and the most granular school ranking possible.31
We complement this result by implementing a Blinder-Oaxaca decomposition of the
finance wage premium. Figure A5 in the online appendix displays both endowment and
coefficient effects for the following determinants of wages in finance: social background,
gender, experience, and talent. Returns to talent appear to explain by far the largest
share of the finance wage premium, whereas talent composition effects are limited.
31
See Table A4 in the online appendix for these robustness checks. The results are also robust to using
the total sample (rather than the trimmed sample), as shown in Table A3 in the online appendix.
22
4.2
Controlling for Treatment Effects with School Fixed Effects
Our result is robust to the inclusion of school fixed effects, which is possible when using
the graduation age as a measure of talent. Columns (4) and (5) of Table 3 report the
regression coefficients when we interact the age at graduation talent measure with our
industry dummies. The specification for both of these regressions includes interaction
terms between school fixed effects and the dummy for working in the finance industry to
better control for potential differentiated effects in the finance industry. We find that,
among alumni from the same school, those who graduate earlier in life are paid relatively
more and that this effect is significantly stronger in finance. The effect is amplified when
we control for industry-year fixed effects (column (5)). This result suggests that neither
treatment effects during school nor school-specific alumni network effects explain our
previous findings, which is consistent with the widely held view in France that most of
the training occurs during the two years of hard work leading up to the selection exam
rather than in the material taught at the schools themselves.
4.3
Controlling for Individual Fixed Effects
We confirm our results by estimating panel data regressions with individual fixed effects
that absorb any time-invariant unobserved characteristics. Returns to talent almost fully
absorb the wage variation when a worker switches into or out of the financial sector.
To include individual fixed effects, we convert our repeated cross-section data into
panel data. We identify unique individuals across time using six sociodemographic variables: year of birth, sex, name of the engineering school, type of specialization and, most
importantly, father’s and mother’s occupations. Identifying individuals along these 7 dimensions minimizes the likelihood of matching observations from different individuals.
This converted panel covers the 2000–2010 period and includes 6,133 individuals who
23
switch sectors over the period, 421 of whom are entering or exiting the finance industry.
We identify the impact of switching sectors on wages using the following regression:
wi,t = αi + β × Ii,t + µ × Dt + λi,t ,
(3)
where αi represents the vector of individual fixed effects, Ii,t is the vector of sector
fixed effects t, and Dt is the vector of year dummies. The results are reported in column
(6) of Table 3.
The 34% wage premium enjoyed by the same individual for working in the finance
industry is close to the level estimated using the cross-section for the same period (29.9%)
and is significantly larger than the premium obtained by workers who enter other sectors.32
To test whether the same individual obtains higher returns to their talent in the finance
industry than in other sectors, we include the interaction of the industry dummy with
talent in this panel specification:
wi,t = αi + β × Ii,t + × Ii,t × T alenti + µ × Dt + λi,t ,
(4)
Column (7) of Table 3 displays the results for this specification. We find that talent
fully absorbs the wage premium realized by a worker who joins the finance industry, and
the coefficient of the finance industry dummy decreases to 9% (but is not significant).
The elasticity of talent is significantly higher in finance than in other sectors. This point
estimate relies on both entry and exit from the finance sector, is robust to including
industry-year fixed effects (column (8)), and provides further evidence that returns to
talent are higher in finance, even when controlling for all unobservable and time-invariant
individual characteristics.
32
This result is consistent with Gibbons and Katz (1992), who find that the wage change experienced
by a typical industry switcher closely resembles the difference in the industry wage differentials estimated
in the cross-section.
24
4.4
Controlling for Job Fixed Effects
We exploit the granularity of our data to ensure that the potential selection of graduates
from top schools into relatively high paying jobs in the industry does not drive our result.
Some occupations in the finance industry, such as traders, are indeed paid much more, on
average, than other jobs.
We reject this endogenous matching explanation by introducing exact job title fixed
effects into equation (1) while restricting the sample to finance workers only. This enables
us to compare, for the same role (e.g., Trader, Quant, Audit, IT), the wages of the alumni
of top and lower ranked schools.33
Our main result is robust to this constrained specification. Column (1) in Table 4
reports the returns to talent for the subsample of individuals for which we possess a job
title without the job titles fixed effects. Moving one notch up our talent scale yields a
7.3% increase in wages for a finance worker, which is close to the level found in our main
specification (column (2) of Table 3). When we include job title fixed effects in column
(2), we still find that returns to talent in the finance industry are more than twice as high
as those in the rest of the economy: moving one notch up our talent scale yields a 4.9%
increase in wages after controlling for job fixed effects. This means that a talented trader,
all else equal, earns significantly more than a less talented one.
INSERT TABLE 4
We complement this analysis by exploring whether returns to talent are higher for
certain job categories or for engineers working in finance outside of France. Columns (3)
and (4) in Table 4 show that returns to talent are significantly higher in front office jobs
(which include Trader, Quant, Structurer, Sales, Asset manager, and Investment Banker)
33
Respondents are asked to give their job title in the 2006–2010 surveys. We manually sort selfdescribed job titles into 9 main job categories for finance workers: back office, support, IT, auditing,
middle office, corporate finance, asset manager, trader, sales, and quant.
25
compared to other jobs in finance (IT, Audit, Middle and Back office, other support
functions). Finally, the left-hand side of Figure 4 displays the estimated returns to talent
for each job category in the finance industry. We observe that returns to talent are more
than twice as high for front office jobs such as Sales, Asset managers, Traders or Quants
than for Auditors or IT workers.
Returns to talent are also stronger for engineers who work in finance outside of France.
Column (5) includes a double interaction term between the talent measure and an indicator variable for working outside of France. The coefficient on this interaction term is
positive and both economically and statistically significant. Returns to talent are indeed
25% higher abroad. The majority of engineers working in finance do so in London or in
New York (more than 50% in our sample), which suggests that returns to talent are even
higher in these geographic zones, perhaps due to more intense competition for talent.
INSERT FIGURE 4
4.5
Increasing Returns to Talent in the Finance Industry and
Allocation Effects
We show that returns to talent have increased over time, which sheds new light on the
increase in the finance premium since the 1980s documented by Philippon and Reshef
(2012). These increasing returns to talent are associated with an increasing share of
talented individuals working in finance.
Columns (1), (2), and (3) of Table 5 report the OLS coefficients of equation (1) over
three periods: the 1980s, the 1990s, and the 2000s. We find that the coefficient on the
interaction term between talent and the finance industry dummy has increased more than
twofold. In the 1980s, a one-notch increase on our talent scale translated into an average
1.8% increase in wages, compared to a 2.8% increase in the finance industry (column (1)).
26
In the 2000s, the same difference in talent generates a 7.6% increase in wages in finance,
compared to a stable 2% increase in the economy at large (column (3)). The residual
of the finance premium, measured by the finance sector dummy, remains stable over the
different periods (columns (1) to (3)). Returns to talent thus increase in line with the
finance wage premium.34
INSERT TABLE 5
The left-hand side of Figure 5 plots the evolution of the share of graduate engineers
from our sample working in finance for the whole sample and for the three top levels of
our talent measure.35 We observe that French graduate engineers have been increasingly
working in finance and that this pattern is significantly more pronounced for the most
talented individuals. While 4% of this group worked in finance in 1986, finance workers
represent more than 10% of this subsample in 2011.36 The share of highly talented
graduates choosing finance, however, might still seem low relative to the significantly
higher returns to talent we observe.
The right-hand side of Figure 5 displays the evolution of the share of graduate engineers
34
A possible explanation for this increase in returns to talent in finance would be a long-term, rigid
supply effect. Thus, the pool of workers identified as talented may not have sufficiently adjusted to
the increase in demand for skills in the finance industry over the last 30 years due to the limited
number of students graduating from top schools. This explanation may not fully explain the increasing trend we observe for two main reasons: first, top schools have been increasing their number of
students over the sample period. Hence, the number of graduating engineers from state engineering schools has increased from 25,000 in 1990 to 40,000 in 2008. The data source can be found at
http : //media.enseignementsup − recherche.gouv.f r/f ile/2009/19/4/RERS20091 19194.pdf . Second,
several studies show that the adjustment costs of supply to demand in the labor market are rather small
over the long run, mainly because shifts in demands are matched by the entry of new workers (Shapiro
(1986), Helwege (1992), Lee and Wolpin (2006)). An alternative explanation for increasing returns to
talent in finance would be that individuals might be more likely to respond to an emailed survey than
to a paper survey if they are working at more skill-intensive tasks and are thus spending more time
on a computer. These talented individuals may then feel more comfortable exaggerating their incomes
online than in paper. However, the survey is anonymous, which limits the incentives to misreport. In
addition, to mitigate this concern, we also estimate the same regressions after winsorizing wages at the
10% threshold for each industry-year. Returns to talent still increase threefold over this period.
35
The pattern is similar if we use only the top level or the top two levels.
36
One concern, as with any survey data, is that the allocation of respondents across industries may
not be perfectly representative of the whole population. However, this concern is mitigated by the fact
that we consider the relative evolution across talent levels and industries. If anything, we should expect
that talented finance workers have reduced incentives to answer the survey, as the opportunity cost of
answering the survey increases.
27
from our sample that work outside of France in finance and in the rest of the economy.
While this share has been increasing for the whole population in line with an increase in
the international mobility of qualified workers, this trend is significantly more pronounced
in finance than in the rest of the economy, especially for the most talented engineers.37
This geographic allocation effect is again consistent with the hypothesis that increasing
returns to talent attract increasingly talented workers. Indeed, returns to talent are even
higher for finance workers working outside of France.
INSERT FIGURE 5
5
Returns to Talent and the Structure of Compensation
We next investigate the relationship between returns to talent and the structure of pay.
Compensation contracts that include a large share of variable pay may be associated
with high returns to talent for several reasons. First, intense competition for talent may
amplify the need for variable pay by increasing the cost of incentivizing talented workers,
either because of their better outside options (Axelson and Bond (2015)) or because the
productivity of their effort is higher. Second, high returns to talent may increase the
need for retention mechanisms. Firms may use variable pay as a sorting mechanism
for attracting and retaining talented workers (Benabou and Tirole (2015), Oyer (2004)).
Third, the high returns to talent we observe in finance may result from better performance
observability than in other industries, which translates into a higher share of variable pay,
as the firm can contract on performance with the worker. We show that variable pay and
returns to talent are closely related; a higher level of talent is associated with a larger share
37
The survey has been e-mailed since 2002, whereas it was previously mailed, which is likely to have
increased participation from alumni working abroad across all sectors. We therefore focus on the evolution
of the share of finance workers working abroad.
28
of variable compensation, and this is even more pronounced in sectors such as finance and
in occupations such as trading in which the returns to talent are especially high.
5.1
Across Industries
Our analysis of variable compensation confirms that variable pay is a key component of
wages in the finance industry, as it is present in 65% of compensation packages in finance
versus 41% in the rest of the economy.
We test whether our talent measure relates to the share of variable compensation
in the finance industry. Column (3) in Table 6 indicates that variable compensation
represents a significantly larger share of total wages in finance than in other sectors and
that more talented workers have a larger share of variable pay. Column (4) includes the
interaction between the talent measure and the finance sector dummy. The coefficient of
the interaction indicates that the effect of talent on the share of variable compensation is
much larger in finance than in the rest of the economy, and the decrease in the coefficient
on the finance dummy, which is divided by three, indicates that this talent effect largely
explains the large share of variable pay in finance.
INSERT TABLE 6
5.2
Across Jobs
Using the detailed information we have on the exact job title of respondents from 2006
to 2009, we explore whether the returns to talent and the structure of compensation
are correlated across jobs within finance. The right-hand side of Figure 4 plots returns
to talent over the share of variable compensation for the main occupation categories in
finance. We observe a strong positive correlation: occupations with the highest returns
to talent also pay the largest share of variable compensation.
29
6
Alternative Channels
This section discusses alternative explanations for our result that are not based on talent
effects.
6.1
Alumni Network Effects
Our results could be driven by alumni network effects rather than by talent. More precisely, the high returns to school ranking we observe in finance might come from alumni
networks being more influential in finance than in the rest of the economy.38
Our specification using age at graduation as a measure of talent and including school
fixed effects, however, controls for school-specific alumni network effects. Assuming that
alumni networks do not discriminate by age of graduation, the fact that our main result
is robust to this specification is not consistent with network effects driving the level of
wages in the finance industry.
In addition, after excluding individuals from the most well-connected schools, i.e., the
Ecole Polytechnique and related schools from our sample, the returns to talent are still
three times higher in the finance industry than in the rest of the economy (column (1) in
Table 7).39
INSERT TABLE 7
Finally, column (7) in Table 2 shows that our talent measure is highly correlated
with the productivity of French graduate engineers who work as academics in the United
States, as measured by the number of citations. Academic paper citations are unlikely to
38
Recent studies on networks insist on their importance in labor market processes, such as hiring,
promotion, and compensation setting (Butler and Gurun (2012), Engelberg et al. (2013) and Shue (2013))
39
Graduates of these schools are over-represented among top executives and CEOs (Kramarz and
Thesmar (2013)). The excluded schools are Ecole Polytechnique, Mines de Paris, Ecole des Ponts,
Supelec, AgroParisTech Grignon, Supaero, INP-ENSEEIHT, Supoptic Orsay, ESPCI Paris, and Chimie
Paris et Telecom Paris. We also exclude Centrale Paris because its selection rate is equivalent.
30
be driven by the network or prestige of the undergraduate engineering school of any of the
authors. Our talent measure is therefore unlikely to capture network effects or prestige
alone.
6.2
Social Background
Our results might be driven by the social backgrounds of graduates if both the share
of students with well-connected parents is higher in top schools and these connections
are particularly valuable in the finance industry. We conduct two distinct tests to rule
out this hypothesis. First, in columns (2) and (3) of Table 7, we restrict our sample to
First Generation students whose parents do not possess university level educations and
are therefore less likely to be well connected. We find that our results are robust to this
subsample; in fact, they are actually strengthened, as the coefficient on the interaction
between talent and finance is significantly higher than for the full sample. Second, in
columns (4) and (5), we restrict our sample to graduates of non-French nationality, following the rationale that their parents are likely to be less integrated into French social
networks. We find that our main result remains robust to this specification, making our
data difficult to reconcile with a social background explanation.
6.3
Compensating Wage Differential
A final alternative explanation would be that higher compensation in finance aims to
offset tougher working conditions or higher income risks. More talented workers deserve
a higher compensation differential because they work relatively harder or because their
health, income or employment are more at risk. However, Philippon and Reshef (2012)
and Oyer (2008)’s estimates of the lifetime pay premium in finance explicitly control for
hours worked, wage risk, career length, and the risks of exiting the finance sector.
31
We nevertheless conduct additional tests to rule out this possibility using data on
job satisfaction and hours worked, controlling for both stress and excessive workloads in
equation (2).40 We use a dummy variable equal to one if a respondent reports suffering
from stress and zero otherwise. We also introduce a variable that indicates whether a
respondent works overtime occasionally, 5 to 10 hours, or more than 10 hours.41 We
find no significant downward impact of these variables on talent returns in the finance
industry. The results are reported in Table A5 in the online appendix.
We employ two strategies to control for unemployment and income risks. We first
observe the fraction of layoffs in the total population of French employees for each sector
as a measure of unemployment risk, and we find that this risk is not higher in finance.42
Second, we restrict our wage regression to the fixed portion of workers’ compensation
packages in columns (1) and (2) of Table 6 and find that finance workers also earn a
premium on fixed pay, which presents low, if any, income risk. In addition, the level of
talent explains the level of fixed compensation in the financial sector.
7
Discussion
This section discusses two main theoretical frameworks that may explain the higher returns to talent observed in the finance industry: competition for talent and rent extraction.
40
We do not control for stress or excessive workloads in our main results, as this information is not
available for the entire sample.
41
Note that self-reported level of stress and job security might be biased. Indeed, if individuals who
work in finance are those who choose risky, high-pressure environments, they might not report different
levels of stress or perceived job security even if a neutral observer would see a large difference. However,
in equilibrium, these individuals should not receive such a large a compensating premium if they do not
suffer from their working conditions. Therefore controlling for workers’ perceptions seems adequate in
our setup.
42
Source: 2009 labor turnover data from the French Ministry of Labor, Employment and
Health. http://travail-emploi.gouv.fr/etudes-recherches-statistiques-de,76/statistiques,78/emploi,82/lesmouvements-de-main-d-oeuvre,272/les-donnees-sur-les-mouvements-de,2268/les-donnees-sur-lesmouvements-de,2633.htmlWe find a negative correlation between wages and industry unemployment
risk: unemployment risk has been constant in the financial sector over the 1999–2011 period (layoff rate
= 1.7%), and the finance sector has one of the lowest layoff rates (whole economy average = 2.9%).
Second, we use a survey question that asks if interviewees experience low job security as an additional
control, which leaves our main result unchanged (see Table A5 in the online appendix).
32
7.1
Labor Market Competition for Talent
In a competitive labor market, firms want to retain talented workers who exhibit high
productivity. There are three main reasons why firm competition for talent in the presence of even small adjustment costs between industries may result in heterogeneous wage
returns to talent among these industries.
First, technology varies across industries, being more skill biased in some. Skill-biased
technologies, such as computers, substitute for routine tasks but complement non-routine
ones, hence increasing the returns to skills and the demand for high-skilled workers (Katz
and Murphy (1992), Autor et al. (1998) and Autor et al. (2003)).
Second, labor market competition also varies across industries. Competition for talent
is higher in industries in which talent is both easily observable and portable across firms
and is generally industry rather than firm specific. The matching of talents to tasks is
more efficient when labor markets are more competitive, resulting in higher returns to
talent.
Third, the scalability of operations plays an important role in talent productivity:
when talent can be easily leveraged, a small difference in talent leads to a high difference
in productivity and, consequently, in wages. Scale effects are high for jobs such as novel
writer or software programmer and low in physically bounded industries in which the level
of physical capital is high, such as for restaurant owners.43
The finance industry ranks high in the three dimensions previously mentioned.44 In
43
A recent example of talent scalability and its potential impact on wages is evening class high school
teachers in South Korea. Talent has always been key in teaching. However, the implementation of online
technologies has led to a shock in teaching scalability, multiplying the productivity of talented teachers.
This scalability effect has led the wages of top teachers to skyrocket, with some earning seven-figure pay
checks. Source: http://online.wsj.com/article/SB10001424127887324635904578639780253571520.html
44
First, the use of information technology is ubiquitous in the finance industry and ranges from realtime databases to powerful in-house risk management and asset pricing software. Philippon and Reshef
(2012) find that the finance industry has relatively high information-technology intensity. In addition,
finance faces much fewer hard-wired technological constraints than does manufacturing, which increases
plasticity and flexibility in financial activities and increases the scope for creative financial innovation.
Second, talent is easily observable and portable across banks. The productivity of a given finance worker
can be quantified by its P&L and can be observed inside and outside the firm at a low cost. This
33
addition, the integration of world capital markets, their deregulation since the 1980s, and
the development of information technologies have reinforced these effects. Philippon and
Reshef (2012) find that deregulation has been associated with an increase in skill intensity,
job complexity, and wages for finance employers. Concerning scalability, Kaplan and
Rauh (2010) estimate that capital per employee at top US securities firms has increased
substantially from $124,000 (in 2004 dollars) in 1972 to $1,789,000 in 2004. They also
observe a twenty-three-fold increase in capital per managing director since the 1970s.
We develop a simple model in the online appendix based on Rosen (1981) and Gabaix
and Landier (2008) in which competition for workers and scalability are associated and
result in higher returns to talent. Introducing adjustment costs between sectors induces
individuals with the same level of talent to earn significantly different wages across industries. These adjustment costs might encompass both differences in preferences among
talented people and switching costs, the latter being likely to be higher for senior positions
or positions requiring a high level of skills.
Our empirical results are consistent with the main predictions of labor market competition theory. First, high returns to talent are associated with high skewness in the
wage distribution, the so-called superstar effect. Second, in the time series, we observe
that returns to talent have increased over the years, which is consistent with an increase
in scalability due to deregulation and financial innovation. Third, returns to talent are
higher in more scalable jobs, such as front office jobs, and in more competitive labor markets, such as in the United Kingdom and the United States. Finally, in the labor market
competition view, variable compensation may be used to attract talented workers when
talent is not perfectly observable (Benabou and Tirole (2015)), which is consistent with
our empirical correlation between returns to talent and pay structure.
facilitates efficient job assignments and capital allocation. Finally, the dematerialized nature of financial
transactions induces low physical constraints and hence high flexibility of capital that can easily flow to
talent.
34
7.2
Extraction of Rents from Employers
Another stream of theoretical literature explores how the finance wage premium may come
from finance workers extracting rents from their employers.45
The main reason why talented workers may be able to extract larger rents from their
employers in finance is that moral hazard problems may be more severe in this industry.46
The exact effort of an employee may be difficult to monitor in innovative financial activities, which are often opaque (Biais et al. (2015), Biais and Landier (2015)). Employee
marginal effort, on the other hand, would be even more critical in finance because of the
large amounts of capital under management. Axelson and Bond (2015) develop a dynamic
model of financial contracting taking into account this severe moral hazard problem in
the finance industry.
The fact that our main result is robust to job title fixed effects suggests that moral
hazard might not be the sole driver of high returns to talent in the finance industry. One
would indeed expect moral hazard to vary more with the type of task and less by the type
of individual accomplishing it.
Moral hazard, however, may interact with competition for talent and thereby result in
higher returns to talent. More talented workers may indeed be more costly to incentivize
when firms compete more for talent because of better outside options. Alternatively, high
powered incentives may serve as a device for firms to attract or retain talented workers,
the latter being more productive in rewardable tasks (Benabou and Tirole (2015)).
This complementary interaction between competition for talent and incentive rents is
also consistent with our empirical results. For example, the positive correlation between
45
This rent extraction is defined within the firm, as opposed to the view where finance workers extract
rents for their employers. In this latter setup, banks will adequately pay workers as a function of their
talent in extracting rents for the bank, which is consistent with the labor market competition theory.
46
A second reason would be that talented finance workers use their skills to capture a larger share
of the pay setting process within banks, for example, through managerial power (Bebchuk and Fried
(2004)). However, it is unclear why banks would consistently hire talented workers who use their skills
to increasingly extract rents from them over time.
35
returns to talent and incentive rents suggest that variable pay is used both to incentivize
talented workers and to attract talent when talent is not perfectly observable (Benabou
and Tirole (2015)).
8
Conclusion
The main contribution of this paper is to show that high and increasing returns to talent
in finance explain the distribution of bankers’ pay and its evolution since the 1980s. To
estimate returns to talent calls for an appropriate measure of talent. We exploit for
this purpose the results of a competitive examination among highly educated candidates
that captures not only cognitive skills but also personality traits such as motivation,
self-discipline, and low cost of effort.
Our results raise questions concerning the possible negative externalities that high
returns to talent in the finance industry might generate. High returns to talent may
lure talented individuals away from other industries (Baumol (1990) and Murphy et al.
(1991)), from regulation (Shive and Forster (2014)) or from socially efficient tasks (Benabou and Tirole (2015), fuel excessively high levels of pay (Glode and Lowery (2015)),
exacerbate bank fragility (Thanassoulis (2012)), or induce inefficient risk-taking (Acharya
et al. (2013)). A related question is whether private returns to talent do not exceed social
returns, if, for instance, banks do not take into account long-term risks.
References
Acharya, V., M. Pagano, and P. Volpin (2013). Seeking Alpha: Excess Risk Taking and Competition for
Managerial Talent. NBER Working Paper Series (18891).
Autor, D., L. F. Katz, and A. Krueger (1998). Computing Inequality: Have Computers Changed the
Labor Market? Quarterly Journal of Economics 113 (4), 1169–1213.
Autor, D., F. Levy, and R. Murnane (2003). The Skill Content of Recent Technological Change: An
Empirical Exploration. Quartlerly Journal of Economics 118 (4), 1279–1333.
36
Axelson, U. and P. Bond (forthcoming, 2015). Wall Street Occupations. Journal of Finance.
Baumol, W. J. (1990). Entrepreneurship: Productive, Unproductive, and Destructive. Journal of Political
Economy 98 (5), 893–921.
Bebchuk, L. and J. Fried (2004). Pay Without Performance: The Unfulfilled Promise of Executive
Compensation. Harvard University Press.
Bell, B. and J. Van Reenen (2013). Extreme Wage Inequality: Pay at the Very Top. American Economic
Review: Papers and Proceedings 103 (3), 153–157.
Bell, B. and J. Van Reenen (2014). Bankers and Their Bonuses. Economic Journal 124 (574), F1–F21.
Benabou, R. and J. Tirole (forthcoming, 2015). Bonus Culture: Competitive Pay, Screening, and Multitasking. Journal of Political Economy.
Biais, B. and A. Landier (2015). Endogenous agency problems and the dynamics of rents. Working
Paper, Toulouse School of Economics.
Biais, B., J.-C. Rochet, and P. Woolley (2015). Dynamics of Innovation and Risk. Review of Financial
Studies 28 (5), 1353–1380.
Bijlsma, M., J. Boone, and G. Zwart (2012). Competition for Traders and Risk. Working Paper .
Bolton, P., T. Santos, and J. A. Scheinkman (2013). Cream Skimming in Financial Markets. Working
Paper .
Bond, P. and V. Glode (2014). The Labor Market for Bankers and Regulators. Review of Financial
Studies 27 (9), 2539–2579.
Boustanifar, H., E. Grant, and A. Reshef (2015). Wages and Human Capital in Finance: International
Evidence, 1970-2007. Working Paper .
Bowles, S., H. Gintis, and M. Osborne (2001). The Determinants of Earnings: A Behavioral Approach.
Journal of Economic Literature 39 (4), 1137–1176.
Butler, A. W. and U. G. Gurun (2012). Educational Networks, Mutual Fund Voting Patterns, and CEO
Compensation. Review of Financial Studies 25 (8), 2533–2562.
Böhm, M., D. Metzger, and P. Strömberg (2015). ”Since you’re so rich, you must be really smart”: Talent
and the Finance Wage Premium. Working Paper .
Cuñat, V. and M. Guadalupe (2005). How does product market competition shape incentive contracts?
Journal of the European Economic Association 3 (5), 1058–1082.
Deming, D. J. (2015). The Growing Importance of Social Skills in the Labor Market. NBER Working
Paper Series (21473).
Engelberg, J., P. Gao, and C. A. Parsons (2013). The Price of a CEO’s Rolodex. Review of Financial
Studies 26 (1), 79–114.
Frydman, C. (2007). Rising Through the Ranks. The Evolution of the Market for Corporate Executives,
1936-2003. Working Paper .
Gabaix, X. and A. Landier (2008). Why Has CEO Pay Increased so Much?
Economics 123 (1), 49–100.
Quarterly Journal of
Gao, H., J. Luo, and T. Tang (forthcoming, 2015). Effects of Managerial Labor Market on Executive
Compensation: Evidence from Job-Hopping. Journal of Accounting and Economics.
Garicano, L. and E. Rossi-Hansberg (2006). Organization and Inequality in a Knowledge Economy.
Quarterly Journal of Economics 121 (4), 1383–1435.
Geerolf, F. (2014). A Static and Microfounded Theory of Zipf’s Law for Firms and of the Top Labor
Income Distribution. Working Paper .
Giannetti, M. and D. Metzger (2013). Compensation and Competition for Talent: Talent Scarcity or
Incentives? Working Paper .
Gibbons, R. and L. Katz (1992). Does Unmeasured Ability Explain Inter-industry Wage Differentials?
The Review of Economic Studies 59 (3), 515–535.
Glode, V. and R. Lowery (Forthcoming, 2015). Compensating Financial Experts. Journal of Finance.
Goldin, C. and L. Katz (2008). Transitions: Career and Family Life Cycles in the Educational Elite.
37
American Economic Review: Papers & Proceedings 98 (2), 363–369.
Greenwood, R. and D. Scharfstein (2013). The Growth of Finance. Journal of Economic Perspectives 27 (2), 3–28.
Guadalupe, M. (2007). Product Market Competition, Returns to Skill, and Wage Inequality. Journal of
Labor Economics 25 (3), 439–474.
Heckman, J. J. (1995). Lessons From the Bell Curve. Journal of Political Economy 103 (5), 1091–1120.
Heckman, J. J. and T. Kautz (2012). Hard Evidence on Soft Skills. Labour economics 19 (4), 451–464.
Helwege, J. (1992). Sectoral Shifts and Interindustry Wage Differentials.
nomics 10 (1), 55–84.
Journal of Labor Eco-
Kaplan, S. and J. D. Rauh (2010). Wall Street and Main Street: What Contributes to the Rise in the
Highest Income? Review of Financial Studies 23 (3), 1004–1050.
Kaplan, S. N. and J. D. Rauh (2013). Family, Education, and Sources of Wealth among the Richest
Americans, 1982-2012. American Economic Review 103 (3), 158–62.
Katz, L. F. and K. M. Murphy (1992). Changes in Relative Wages, 1963–1987: Supply and Demand
Factors. Quarterly Journal of Economics 107 (1), 35–78.
Kramarz, F. and D. Thesmar (2013). Social Networks in the Boardroom. Journal of European Economic
Association 11 (4), 780–807.
Lee, D. and K. I. Wolpin (2006). Intersectoral Labor Mobility and the Growth of the Service Sector.
Econometrica 74 (1), 1–46.
Lemieux, T., W. B. MacLeod, and D. Parent (2009). Performance Pay and Wage Inequality. Quarterly
Journal of Economics 124 (1), 1–49.
Levy, F. and P. Temin (2007). Inequality and Institutions in 20th Century America. Number 13106.
Murphy, K., A. Shleifer, and R. Vishny (1991). The Allocation of Talent: Implications for Growth.
Quarterly Journal of Economics 106 (2), 503–530.
Murphy, K. J. and J. Zábojnı́k (2004). CEO Pay and Appointments: A Market-Based Explanation for
Recent Trends. American Economic Review 94 (2), 192–196.
Ors, E., F. Palomino, and E. Peyrache (2013). Performance Gender-Gap: Does Competition Matter?
Journal of Labor Economics 31 (3), 443–499.
Oyer, P. (2004). Why Do Firms Use Icentives That Have No Icentive Effects? Journal of Finance 69 (4),
1619–1650.
Oyer, P. (2008). The Making of an Investment Banker: Stock Market Shocks, Career Choice, and Lifetime
Income. Journal of Finance 63 (6), 2601–2628.
Philippon, T. and A. Reshef (2012). Wages and Human Capital in the U.S. Finance Industry: 1909-2006.
Quartlery Journal of Economics 127 (4), 1551–1609.
Piketty, T. and E. Saez (2003). Income Inequality in the United States, 1913-1998. Quarterly Journal of
Economics 118 (1), 1–41.
Piketty, T. and E. Saez (2006). The Evolution of Top Incomes: A Historical and International Perspective.
American Economic Review 96 (2), 200–205.
Rosen, S. (1981). The Economics of Superstars. American Economic Review 71 (5), 845–858.
Shapiro, M. D. (1986). The dynamic demand for capital and labor.
nomics 101 (3), 513–542.
Quarterly Journal of Eco-
Shive, S. and M. Forster (2014). The Revolving Door for Financial Regulators. Working Paper .
Shu, P. (2013). Career Choice and Skill Development of MIT Graduates: Are the “Best and Brightest”
Going into Finance? Working Paper .
Shue, K. (2013). Executive Networks and Firm Policies: Evidence from the Random Assignment of MBA
Peers. Review of Financial Studies 26 (6), 1401–1442.
Thanassoulis, J. (2012). The Case for Intervening in Bankers’ Pay. Journal of Finance 67 (3), 849–895.
38
A
Figures
Figure 1. Selection Process of French Engineering Schools
High School - Science Major
⇓
Preparatory School (2 years, Selection rate: 15% )
Subject
# Hours
(per Week)
# Exams
(per month)
12
8
2
2
2 to 4
2
4
2
0
1
2
0
Mathematics
Physics and Chemistry
Industrial Science
Literature
Foreign Language
Programming
⇓
National Competitive Exam
Written Competitive Exam
Subject
Oral Competitive Exam
Subject
Coefficient
Mathematics 1
Mathematics 2
Physics
Industrial Science
Literature
Foreign Language
Computer Science
Coefficient
Mathematics 1
Mathematics 2
Physics
Industrial Science
Literature
Foreign Language
Chemistry
Sports
8
7
6
6
6
6
4
16
16
20
15
8
8
9
5
⇓
Final Ranking
60
40
0
20
Selection Rate, in %
80
100
Distribution of Engineering Schools by Selection Rate
1
2
3
4
5
6
7
8
9
10
School Rank
Note: This figure summarizes the selection process to enter in French Engineering Schools and displays the resulting
distribution of engineering school selectivity by level of talent. French engineering schools, or “Grandes Ecoles”, select
students for admission based on their national ranking in a competitive written and oral exam. Schools are sorted on their
selection rate, measured as the ratio of the marginal student’s rank in the national competitive exam to the total number
of competing students.
39
0
Finance Wage Premium, in %
20
40
60
80
Figure 2. Evolution of the Finance Wage Premium by Percentile of the Wage
Distribution
1980s
1990s
10th Percentile
50th percentile
2000s
90th percentile
Note: This figure plots the evolution of the coefficient of the financial sector dummy in quantile regressions estimated at
the 10th, 50th, and 90th percentiles of the wage distribution, in which the dependent variable is the log of the yearly gross
wage. There are 48 industry dummies, with the sum of all industry dummy coefficients being constrained to zero. Each
regression also controls for education, gender, marital status, occupation, firm type, firm size, hierarchical responsibilities,
working abroad, working in the Paris area, experience, experience squared, and experience cubed.
40
100000
80,000
60,000
40,000
Yearly Gross Wage (in Euros)
120000
Figure 3. Predicted Wages by School Rank and Sector
10
9
8
7
6
5
4
3
Whole economy
Chemistry
Consulting
Oil
2
1
Finance
Note: This figure displays the predicted yearly gross wage calculated from the estimation of an OLS regression at the
different levels of our talent scale, with average values for all other variables. The dependent variable in the estimation is
the log of the yearly gross wage, and is estimated over the 2004-2011 period for five different samples: the whole economy
(124,433 observations), and the chemistry (2,752 observations), oil (717 observations), consulting (3,773 observations), and
finance (3,431 observations) industries. The model includes a female dummy, a married dummy, a female × married dummy,
a Paris area dummy, a working abroad dummy, six country dummies, experience level squared and cubed, four hierarchic
responsibility dummies, nine occupation dummies, four firm size dummies, and four firm type dummies.
41
.07
.07
Figure 4. Returns to Talent and the Structure of Pay across Jobs in Finance
Quant
Trader
Returns to Talent
.05
Returns to Talent
.03
.04
.05
.06
.06
Sales
Asset Manager
.02
Corporate Finance
ra
.03
s
le
Sa
r
Q
ua
nt
er
de
ag
Tr
a
ce
an
tM
se
As
te
Fi
n
an
e
g
O
ffi
c
e
dl
id
po
or
IT
C
IT
tin
di
Au
M
Auditing
10
R
es
to
ft
he
Ec
on
om
y
.01
.04
Middle Office
20
30
40
Share of variable compensation in %
50
Note: This figure displays the estimated returns to talent for each job category in the Finance industry, and plot these
returns over the share of variable compensation on the right hand side of the figure. Self described job titles of individuals
from the 2000-2010 surveys have been manually sorted into job categories. Returns to talent are the coefficients on the
interaction terms between our talent measure and job category indicator variable, in OLS regressions where the dependent
variable is the log of the yearly gross wage. The model includes a female dummy, a married dummy, a female × married
dummy, a Paris area dummy, a working abroad dummy, experience level squared and cubed, four hierarchic responsibility
dummies, and four firm size dummies.
42
2
40
30
0
10
20
Share of Alumni Working Abroad (in \%)
6
4
Share of Alumni in Finance (in \%)
50
8
Figure 5. Alumni Allocation in Finance by Talent Level and Geographical
Area
1985
1985
1990
1995
2000
2005
2010
1995
2000
2005
2010
Year
Year
Top Talent
1990
Finance*Top Talent
All Talent Levels
Finance
Rest of the Economy
Note: The left-hand side of this figure plots the evolution of the share of graduate engineers from our sample working in
finance, for the total sample and for the three top levels of our talent measure. The right-hand side of this figure displays
the evolution of the share of graduate engineers from our sample that work outside of France, for the whole sample, for
engineers working in finance, and for top talented engineers working in finance.
43
B
Tables
Table 1. Summary Statistics
1980s
1990s
2000s
Sample Size
Average number of observations per survey
Number of surveys
Total number of observations
Response rate (%)
Coverage of total population of French engineers (%)
20,805
3
62,415
21
9
15,088
4
60,353
17
7.1
17,776
7
124,433
Nd
6.2
Compensation (in 2005 constant euros)
Mean yearly gross wage
90th percentile
99th percentile
Standard deviation
62,137
99,718
146,253
27,073
62,625
101,964
169,870
31,827
57,983
95,598
186,438
39,086
Engineers per sector (%)
Finance
Consulting
Oil
Chemical
1.9
0.0
3.1
3.6
2.3
1.5
1.8
3.8
3.5
3.6
0.7
2.6
Demographics
Mean age
Percent female
Percent married
Foreigners
First Generation
38.4
6.1
77.7
-
38.2
11.9
73.6
-
35.1
15.3
77.2
8.6
11.8
Work location
Percent working outside of France
Percent working in the Paris area
2.6
46.9
4.1
42.4
12.1
39.3
Career
Mean experience (years)
Percent team manager
Percent department head
Percent top executive
14.6
32.1
15.9
6.5
13.6
25.2
19.2
11.3
11.9
21.4
17.7
7.1
This table reports summary statistics for the main compensation and demographic variables in our dataset.
1980s = graduates from the 1983, 1986, and 1989 surveys; 1990s = graduates from the 1992, 1995, 1998,
and 2000 surveys; 2000s = graduates from the 2004, 2005, 2006, 2007, 2008, 2010, and 2011 surveys. Source:
IESF Compensation Survey.
44
Table 2. Measuring Talent
School
Rank
Recruitment
Level
Graduates
#
Schools
% Early
Acceptance
# Citations
(Academics)
% Share
(5)
(6)
(7)
(1)
(2)
(3)
Number
(4)
1
2
3
4
5
6
7
8
9
10
Top 2%
Top 5%
Top 10%
Top 15%
Top 30%
Top 40%
Top 50%
Top 60%
Top 80%
100%
1
3
7
5
7
8
14
21
45
85
6,173
12,868
20,119
12,236
12,182
11,468
46,676
20,747
36,615
47,762
2.7
5.7
8.9
5.4
5.4
5.1
20.6
9.1
16.1
21.1
36.0
21.2
14.2
12.9
17.1
11.4
13.0
8.8
11.1
10.2
3,174∗∗∗
1,897∗∗∗
1,623∗∗
1,147∗∗
-
Total
-
196
226,846
100.0
-
-
This table reports summary statistics for each level of our talent measure School Rank. This talent measure
takes a value from 1 to 10 and sorts schools based on their selectivity rate. French engineering schools, or
“Grandes Ecoles”, select students for admission based on student national ranking in a competitive written
and oral exam. Recruitment level (column (2)) is the position of the marginal student for each school in the
national ranking. Column (3) reports the number of schools for each level of our talent measure. Columns
(4) and (5) give the number and share of students for each level of talent. Column (6) reports the share of
respondents that are admitted in an engineering school early (at least one year ahead). Column (7) gives the
predicted number of citations (after 9.7 years of experience) of academics that graduated from a school of
the corresponding rank. The predicted values result from an OLS regression where the dependent variable is
the 1% winsorized number of citations in Google Scholars. The sample consists of the 98 graduates from top
engineering schools working in US universities. The control variables are gender, research area, experience,
experience squared and experience cubed. Standard errors are clustered at the school level.
45
Table 3. Heterogenous Wage Returns to Talent across Industries
Log(Wage)
OLS
Talent Measure
Finance
11-School Rank
(1)
(2)
0.248***
(0.033)
0.024
(0.026)
Talent × Finance
Consulting
0.044***
(0.006)
0.139***
(0.012)
Talent × Consulting
Oil
0.128***
(0.010)
Talent × Oil
Chemistry
Talent × Chemistry
(3)
(4)
(5)
0.007
(0.025)
0.046***
(0.005)
0.032**
(0.013)
0.020***
(0.003)
0.020
(0.015)
0.089***
(0.011)
0.004
(0.017)
0.052
(0.032)
0.053**
(0.022)
0.021***
(0.003)
0.019***
(0.003)
0.019***
(0.003)
0.025***
(0.006)
0.025***
(0.005)
Individual FE
-
-
-
-
-
School FE
-
-
-
Yes
Individual Controls
Yes
Yes
Yes
Year FE
Yes
Yes
-
(8)
0.052**
(0.021)
0.057
(0.065)
0.006
(0.015)
0.005
(0.015)
0.102
(0.112)
0.001
(0.019)
0.107***
(0.041)
0.002
(0.013)
R2
0.077
(0.102)
0.001
(0.016)
0.003
(0.013)
Observations
0.331***
(0.066)
0.104*
(0.056)
-0.004**
(0.002)
Industry-Year FE
(7)
0.021
(0.016)
-0.004**
(0.002)
Talent
(6)
0.080**
(0.038)
0.139**
(0.059)
-0.005
(0.003)
11-School Rank
0.047***
(0.015)
0.042
(0.029)
0.155***
(0.019)
-0.005
(0.003)
0.072***
(0.007)
Graduation Age
(# Years Ahead)
0.049***
(0.017)
0.020***
(0.003)
Panel
-0.001
(0.019)
0.105
(0.071)
0.001
(0.017)
-0.000
(0.017)
Yes
Yes
Yes
Yes
-
-
-
Yes
Yes
-
-
-
Yes
Yes
Yes
Yes
Yes
Yes
-
Yes
-
Yes
-
-
Yes
198,886
198,886
198,886
52,332
52,332
62,718
62,718
62,718
0.698
0.701
0.703
0.557
0.561
0.950
0.951
0.951
This table reports the coefficient of OLS regressions, where the dependent variable is the log of yearly gross
wage. All specifications include dummies for working in the oil, finance, chemistry, and consulting industries.
In columns (1), (2), (3), (6), (7) and (8), Talent is equal to 11-School Rank, with School Rank based on
the ranking of the marginal student in the national competitive exam, as defined in table 2. In column (4)
and (5), Talent is equal to 26 - Age at Graduation. The average age at graduation is 23 years old. Highly
performing students graduate earlier on average because they often skipped a year during primary school,
whereas less talented students often repeat years during prep school to improve their result at the national
competitive exam. In columns (6), (7) and (8), the sample is restricted to the 15,256 individuals that
are uniquely identified and tracked over the 2000-2010 period through their demographic characteristics.
Column (4) and (5) includes school fixed effects, and columns (6), (7) and (8) include individual fixed effects.
All equations include year dummies, a female dummy, a married dummy, a female × married dummy, a Paris
area dummy, a working abroad dummy, experience level squared and cubed, four hierarchic responsibility
dummies, nine occupation dummies, four firm size dummies, and four firm type dummies. Columns (3),
(5) and (8) also include year-industry fixed effects. Standard errors are clustered at the school level and
reported in brackets, * p<0.10, ** p<0.05, *** p<0.01.
46
Table 4. Returns to Talent and Jobs in Finance
Log(Wage)
Sample
Talent
Finance Workers
(1)
(2)
(3)
(4)
0.073***
(0.007)
0.049***
(0.005)
0.053***
(0.005)
0.045***
(0.004)
0.487***
(0.035)
0.365***
(0.061)
Front Office
Talent × Front Office
0.067***
(0.007)
0.022**
(0.009)
Talent × Working Abroad
Job Fixed Effects
(5)
0.018**
(0.009)
-
Yes
-
-
-
Individual Controls
Yes
Yes
Yes
Yes
Yes
Year Fixed Effects
Yes
Yes
Yes
Yes
Yes
2,399
2,399
2,399
2,399
2,399
0.512
0.622
0.593
0.595
0.513
Observations
R
2
This table reports the coefficient of OLS regressions, where the dependent variable is the log of the yearly
gross wage. Yearly gross wage includes both fixed and variable compensation. The sample is restricted
to the 2,399 workers in the finance industry who provide their exact job title. Columns (2) includes job
category fixed effects. Self described job titles of individuals from the 2000-2010 surveys have been manually
sorted into job categories, including IT, Auditing, Middle Office, Corporate Finance, Asset Manager, Sales,
Trader and Quant. Columns (3) and (4) include a indicator variable for front office jobs, which include
traders, quants, sales, investment bankers, and asset managers, and in column (4) this indicator variable is
interacted with our talent measure. Column (5) include a dummy variable Working Abroad equal to one
if the alumni works outside of France, which is interacted with our talent measure. All equations include
year dummies, a female dummy, a married dummy, a female × married dummy, a Paris area dummy, a
working abroad dummy, experience level squared and cubed, four hierarchic responsibility dummies, and
four firm size dummies. Standard errors are clustered at the school level and reported in brackets, * p<0.10,
** p<0.05, *** p<0.01.
47
Table 5. Increasing Wage Returns to Talent in the Finance Industry
Log(Wage)
S1980
(1)
S1990
(2)
S2000
(3)
Finance
0.016
(0.021)
0.011
(0.026)
0.020
(0.026)
Talent × Finance
0.010**
(0.004)
0.024***
(0.004)
0.056***
(0.005)
Talent
0.018***
(0.002)
0.018***
(0.003)
0.020***
(0.003)
Individual Controls
Yes
Yes
Yes
Year Fixed Effects
Yes
Yes
Yes
41,731
52,932
104,223
0.713
0.715
0.694
Observations
R
2
This table reports the coefficient of an OLS regression over three samples: S1980 = 1986 and 1989 surveys
(Column (1)); S1990 = 1992, 1995, 1998, and 2000 surveys (Column (2)); and S2000 = 2004, 2005, 2006,
2007, 2008, 2010, and 2011 surveys (Column (3)). The dependent variable is the log of the yearly gross
wage. Yearly gross wage includes both fixed and variable compensation. Talent (which takes a value from
1 to 10) is equal to 11-School Rank, with School Rank based on the ranking of the marginal student in the
national competitive exam, as defined in table 2. All equations include year dummies, a female dummy, a
married dummy, a female × married dummy, a Paris area dummy, a working abroad dummy, experience
level squared and cubed, four hierarchic responsibility dummies, nine occupation dummies, four firm size
dummies, and four firm type dummies. Standard errors are clustered at the school level and reported in
brackets, * p<0.10, ** p<0.05, *** p<0.01.
48
Table 6. Returns to Talent and Compensation Structure
Fixed Compensation
Log (Fixed Wage)
Finance
(1)
(2)
(3)
(4)
0.045***
(0.012)
-0.002
(0.019)
0.863***
(0.041)
0.250***
(0.075)
Talent × Finance
Talent
Variable Share
Log(1 + Share)
0.009***
(0.003)
0.118***
(0.012)
0.024***
(0.002)
0.023***
(0.002)
0.053***
(0.003)
0.045***
(0.003)
Individual Controls
Yes
Yes
Yes
Yes
Year Fixed Effects
Yes
Yes
Yes
Yes
52,777
52,777
52,777
52,777
0.413
0.413
0.134
0.136
Observations
R
2
This table reports the coefficient of OLS regressions, where the dependent variable is the log of the yearly
fixed wage in columns (1) and (2), and of the share of variable wage in columns (3) and (4). The sample is
restricted to the period 2000 to 2011 for which our data includes information on the structure of pay. All
equations include year dummies, a female dummy, a married dummy, a female × married dummy, a Paris
area dummy, a working abroad dummy, experience level squared and cubed, four hierarchic responsibility
dummies, nine occupation dummies, four firm size dummies, and four firm type dummies. Standard errors
are clustered at the school level and reported in brackets, * p<0.10, ** p<0.05, *** p<0.01.
49
Table 7. Controlling for Network and Social Background Effects
Log(Wage)
Sample
Finance
No-X
Schools
First
Generation
Foreigners
(1)
(2)
(3)
(4)
0.047
(0.032)
0.340***
(0.053)
0.020
(0.059)
0.471***
(0.057)
(5)
0.175*
(0.097)
Talent × Finance
0.037***
(0.010)
Talent
0.016***
(0.003)
0.020***
(0.003)
0.018***
(0.002)
0.023***
(0.004)
0.020***
(0.004)
Individual Controls
Yes
Yes
Yes
Yes
Yes
Year Fixed Effects
Yes
Yes
Yes
Yes
Yes
178,377
14,488
14,488
1,399
1,399
0.689
0.660
0.665
0.561
0.566
Observations
R
2
0.074***
(0.013)
0.047***
(0.012)
This table reports the coefficient of OLS regressions, where the dependent variable is the log of the yearly
gross wage. In column (1) the sample is restricted to schools that are not related to Ecole Polytechnique,
the leading French Engineering school (The 14 excluded schools are Ecole Polytechnique, Mines de Paris,
Ecole des Ponts, Supelec, AgroParis-Tech Grignon, Arts et Metiers Paris-Tech, Supaero, INP-ENSEEIHT,
Ensta, Supoptic Orsay, ESPCI Paris, Chimie Paris, and Telecom Paris). In columns (2) to (3) the sample
is restricted to ”first generation” students, whose parents do not have college education (the information
is available from 2000 to 2010). In columns (3) to (4), the sample is restricted to individuals born outside
France (the information is available from 2000 to 2010). Finally, in columns (6) and (7), the sample is
restricted to individuals working outside France. All equations include year dummies, a female dummy, a
married dummy, a female × married dummy, a Paris area dummy, a working abroad dummy, experience
level squared and cubed, four hierarchic responsibility dummies, nine occupation dummies, four firm size
dummies, and four firm type dummies. Standard errors are clustered at the school level and reported in
brackets, * p<0.10, ** p<0.05, *** p<0.01.
50
For Online Publication
Appendix A -
List of Main Variables
Selection rate: the ratio of the rank of the last admitted candidate to the total number of applicants.
Graduation age: the age at which a student obtains the “Engineer” degree; in France, a student who
has neither skipped nor repeated a year of schooling usually graduates at 23 years of age.
Predicted wage: the wage obtained when predicting wages using the coefficients of the main equation.
Early graduation: an indicator variable for graduating earlier than the standard age (23 years old).
Top manager : an indicator variable for holding a top management position, defined in the survey by
being on the executive committee.
Finance: an indicator variable for working in the financial sector, which includes banks, investment
funds, and insurance companies.
Wage: the gross annual salary of a given engineer, as disclosed in the alumni survey. It includes both
fixed and cash variable compensation.
Variable compensation: the annual amount of variable compensation, disclosed in a specific question
on the survey.
Job title: the exact occupation within finance (e.g., trader, risk manager, investment banker).
School rank : the level of selectivity of a given engineering school within ten categories (see table ?? in
the appendix for the list of schools by level of selectivity).
X-schools: schools affiliated with the top French engineering school, Ecole Polytechnique.
51
Appendix B -
Methodology for Selection Rate Calculation
We compute a school’s selection rate by dividing the rank in the national exam of the last admitted
student by the total number of enrolled students nationwide. We use information from the end of that
sample period, as the level of school selectivity is strongly persistent. Our results are robust to using the
average over the period. For prominent schools, namely “Ecole Polytechnique”, all “Ecole Centrales”,
“Mines”, “Ponts et Chaussées”, “Supelec”, “Supaero” and “Telecom Paris”, we take the rank of the last
admitted student as given. As an example, in 2012, the marginal student in the mathematics option in
Ecole Polytechnique is ranked 124th , and 8,343 students take the national exam. Hence, the selection
rate of Ecole Polytechnique is 1.5%. Because some students self-select and do not apply to the lower
ranked schools, the rank of the marginal student for the other schools is biased upward. We therefore
adjust the rank of the last admitted students for these schools by adding to the marginal student rank
the number of students that do not apply. This calculation therefore assumes that the students that do
not apply would be admitted if they do. Back to our example, the rank of the last admitted student in
Enac Toulouse in 2012 is 1,645th in the mathematics option. Given that only 7,094 students applied to
this school out of a total of 8,343 enrolled students nationwide, the adjusted rank of the last admitted
students is 2,894th = 1,645+ (8,343-7094), which gives a selection rate of 34.7% (2894/8343).
A smaller group of schools admit students directly after the Baccalauréat, and not through the
national competitive exam following preparatory school. For this subgroup, we measure the selectivity of
the Engineering school by using the average Baccalauréat grade of their admitted students. We allocate
these schools across groups 7 to 10, where the average Baccalauréat grade of admitted students from
other schools is comparable. Schools allocated to group 7 have an average Baccalauréat grade around
16/20, whereas the ones allocated to group 10 are around 12/20.47
47
Our results hold when excluding these schools (see Table A4 in the online appendix for more details.)
52
Appendix C -
Figures
0
10,000
Yearly Gross Wage
20,000 30,000 40,000
50,000
Figure A1. Starting Salaries: French Engineering School Alumni Survey vs.
Towers Watson Survey
Chemicals Finance
Industry
Rank 1
Alumni Survey
Rank 2
Rank 3
Rank 4
Employer Survey (Towers Watson)
Note: This graph plots the median yearly gross wage across sectors and levels of talent in 2008 and in 2010 for workers of
less than four years of experience. Sources are the IESF alumni survey and the employer survey from Towers Watson.
53
5
10
Finance Premium in %
15
20
25
30
35
Figure A2. The Finance Wage Premium Evolution
1985
1990
1995
2000
Year of the survey
2005
2010
Note: The figure displays the evolution of the coefficient of the financial sector dummy in OLS regressions estimated over the
1983-2011 period, in which the dependent variable is the log of the yearly gross wage. All equations include year dummies,
a female dummy, a married dummy, a female × married dummy, a Paris area dummy, a working abroad dummy, experience
level squared and cubed, four hierarchic responsibility dummies, nine occupation dummies, four firm size dummies, and
four firm type dummies.
54
Figure A3. Evolution of the Wage Premium in the Finance, Chemistry, Oil
and Consulting Industries at the 10th , 50th and 90th percentiles of the wage
distribution
Oil
Wage Premium in %
10203040506070
Wage Premium in %
10 20 30 40 50 60 70
Finance
1980
1990s
year
2000s
1980
2000s
Consulting
Wage Premium in %
10203040506070
Wage Premium in %
10203040506070
Chemistry
1990s
year
1980
1990s
year
2000s
1980
0.1q
0.9q
1990s
year
2000s
0.5q
Note: Each graph displays the evolution of industry dummy coefficients in quantile regressions estimated over three periods:
the 1980s (1983, 1986 and 1989 surveys), the 1990s (1992, 1995, 1998 and 2000 surveys, 53,680 observations) and the 2000s
(2005, 2006, 2007, 2009 and 2011 surveys, 96,611 observations) for q=0.1, q=0.5 and q=0.9. The four considered industries
are the Finance, the Oil, the Chemical and the Consulting industries, which are the four highest-paying industries over the
period. Each regression also controls for education, gender, marital status, occupation, firm type, firm size, hierarchical
responsibilities, working abroad, working in Paris area, experience, experience squared and experience cubed.
55
20
0
10
Frequency, in %
30
40
Figure A4. Distribution of Age at Graduation
21
22
23
24
25
26
Graduation Age
Note: This figure plots the distribution of graduation age across the survey sample, which maps into age at entry. Heterogeneity results mainly from some students skipping years before high school, while others repeat a year, typically the
second year of preparatory class to improve their performance to the national competitive exam.
56
Figure A5. Blinder-Oaxaca Decomposition of the Finance Wage Premium
0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 First Genera3on Foreigner Gender Experience Talent Other Total -­‐0.05 Endowment Coefficient Note: This figures plots the results of the Blinder-Oaxaca decomposition of the gap in the log of the yearly gross wage
between finance and non finance workers.
57
Appendix D -
Tables
Table A1. Standard Determinants of Wages (Controls)
Log(Wage)
(1)
Female
Age
Married
Experience (years)
Experience2
Experience3
Paris Area
Outside France
Talent
Hierarchical Responsabilities: Team Manager
Hierarchical Responsabilities: Department Head
Hierarchical Responsabilities: Top Executive
Occupation: Production
Occupation: IT
Occupation: Sales
Occupation: Office Work
Occupation: Head Office
Firm Size: 20 to 500 employees
Firm Size: 500 to 2000 employees
Firm Size: >2000 employees
Firm Type: Private Sector
Firm Type: State Firm
Firm Type: Administration
Firm Type: Other
Year Fixed Effects
Observations
R2
-0.073***
(0.002)
0.009***
(0.000)
0.035***
(0.001)
0.054***
(0.001)
-0.002***
(0.000)
0.000***
(0.000)
0.115***
(0.001)
0.323***
(0.002)
0.023***
(0.000)
0.076***
(0.001)
0.204***
(0.002)
0.322***
(0.003)
0.002
(0.002)
-0.008***
(0.002)
0.064***
(0.002)
0.112***
(0.003)
0.156***
(0.004)
0.081***
(0.002)
0.127***
(0.002)
0.159***
(0.002)
0.065***
(0.004)
0.020***
(0.004)
-0.178***
(0.004)
-0.090***
(0.008)
Yes
198,886
0.687
This table reports coefficients of OLS regressions over the total sample. The dependent variable is the log
of the yearly gross wage. The explanatory variables include all the controls used in the paper.
58
Table A2. The Finance Premia
MEAN
(1)
10TH
(2)
50TH
(3)
90TH
(4)
1983 Premia
0.080
(0.014)
0.022
(0.022)
0.057
(0.013)
0.091
(0.022)
1986 Premia
0.032
(0.011)
-0.002
(0.019)
0.029
(0.012)
0.029
(0.019)
1989 Premia
0.090
(0.011)
0.034
(0.017)
0.072
(0.012)
0.141
(0.016)
1992 Premia
0.086
(0.012)
0.045
(0.021)
0.058
(0.010)
0.081
(0.012)
1995 Premia
0.120
(0.017)
0.050
(0.033)
0.090
(0.015)
0.177
(0.025)
1998 Premia
0.131
(0.013)
0.035
(0.018)
0.074
(0.013)
0.169
(0.022)
2000 Premia
0.163
(0.014)
0.021
(0.019)
0.076
(0.012)
0.344
(0.026)
2004 Premia
0.250
(0.015)
0.071
(0.020)
0.126
(0.016)
0.579
(0.022)
2005 Premia
0.272
(0.013)
0.053
(0.018)
0.173
(0.012)
0.589
(0.019)
2006 Premia
0.320
(0.011)
0.082
(0.015)
0.163
(0.009)
0.740
(0.017)
2007 Premia
0.320
(0.010)
0.080
(0.015)
0.192
(0.009)
0.740
(0.014)
2008 Premia
0.231
(0.011)
0.068
(0.015)
0.125
(0.010)
0.479
(0.018)
2010 Premia
0.287
(0.012)
0.109
(0.016)
0.190
(0.012)
0.622
(0.020)
2011 Premia
0.301
(0.014)
0.096
(0.017)
0.219
(0.011)
0.655
(0.019)
1.109
0.329
0.659
2.837
Trend Estimate
This table, which replicates Table 6 in Bell and Van Reenen (2010), reports coefficients of annual OLS
(column (1)) and quantile regressions for q = 0.1 (column (2)), q = 0.5 (column (3)), and q = 0.9 (column
(4)). The dependent variable is the log of the yearly gross wage. All equations include a female dummy, a
married dummy, a female × married dummy, a Paris area dummy, school fixed effects, a working abroad
dummy, experience level squared and cubed, four hierarchic responsibility dummies, nine occupation dummies, four firm size dummies, and four firm type dummies. Standard errors are reported in parentheses.
Trend estimates are multiplied by 100 and adjusted by the number of years so as to be interpretable as the
% relative annual wage increase for finance workers.
59
Table A3.
Sample
Heterogenous Wage Returns to Talent across Industries - Full
Log(Wage)
Talent Measure
Finance
11-School Rank
(1)
(2)
0.253***
(0.036)
0.012
(0.028)
Talent × Finance
Consulting
0.145***
(0.015)
Talent × Consulting
Oil
0.124***
(0.010)
Talent × Oil
Chemistry
(3)
Talent × Chemistry
(4)
(5)
0.002
(0.026)
0.047***
(0.006)
0.049***
(0.005)
0.034**
(0.013)
0.036**
(0.018)
0.083***
(0.026)
0.037
(0.027)
0.024***
(0.004)
0.024***
(0.004)
0.020
(0.014)
0.147***
(0.018)
-0.004
(0.003)
0.071***
(0.008)
Graduation Age
(# Years Ahead)
0.048***
(0.015)
0.021
(0.015)
0.142**
(0.058)
-0.004
(0.003)
0.092***
(0.011)
0.002
(0.017)
-0.001
(0.017)
0.045
(0.033)
-0.005***
(0.002)
-0.005***
(0.002)
0.007
(0.015)
0.007
(0.016)
0.022***
(0.003)
0.020***
(0.003)
0.020***
(0.003)
0.026***
(0.007)
0.026***
(0.006)
-
-
-
Yes
Yes
Individual Controls
Yes
Yes
Yes
Yes
Yes
Year FE
Yes
Yes
Yes
Yes
Yes
-
-
Yes
-
Yes
201,324
201,324
201,324
52,711
52,711
0.673
0.676
0.678
0.549
0.552
Talent
School FE
Industry-Year FE
Observations
R
2
This table reproduces Table 3 in the paper on the full survey sample, without trimming compensation at
the top 1% of the distribution at the industry-year level.
60
Table A4. Alternative Talent Measures
Log(Wage)
Talent Measure
Finance
1-Selection Rate
(1)
(2)
(3)
0.284***
(0.038)
0.044
(0.038)
0.287***
(0.039)
Talent × Finance
Consulting
0.366***
(0.074)
0.161***
(0.014)
Talent × Consulting
Oil
0.122***
(0.009)
0.163***
(0.014)
0.140***
(0.020)
0.065***
(0.007)
0.079***
(0.013)
0.041
(0.042)
0.076***
(0.021)
0.001***
(0.000)
0.124***
(0.009)
-0.025
(0.030)
Talent × Chemistry
Talent
0.069***
(0.020)
(4)
0.003***
(0.001)
0.151***
(0.036)
Talent × Oil
Chemistry
Granular School Rank
(by selection rate)
0.132***
(0.021)
-0.000
(0.000)
0.063***
(0.007)
-0.025
(0.017)
0.075***
(0.013)
-0.000
(0.000)
0.162***
(0.028)
0.146***
(0.026)
0.001***
(0.000)
0.001***
(0.000)
Individual Controls
Yes
Yes
Yes
Yes
Year Fixed Effects
Yes
Yes
Yes
Yes
152,487
152,487
152,487
152,487
0.698
0.700
0.696
0.698
Observations
R2
This table reports the coefficient of standard OLS regressions. The dependent variable is the log of the
yearly gross wage. The oil, finance, chemistry, and consulting industries have a dummy variable. Talent
is equal to 1 - the school selection rate at the national competitive exam in columns (1) and (2), and to
the exact rank of each school according to this selection rate (no categories) in columns (3) and (4). All
equations include year dummies, a female dummy, a married dummy, a female × married dummy, a Paris
area dummy, a working abroad dummy, experience level squared and cubed, four hierarchic responsibility
dummies, nine occupation dummies, four firm size dummies, and four firm type dummies. Standard errors
are clustered at the school level and in parentheses, * p<0.10, ** p<0.05, *** p<0.01.
61
Table A5. Working Conditions
Log(Wage)
(1)
(2)
(3)
Finance
0.313***
(0.043)
0.034
(0.027)
0.032
(0.026)
0.029
(0.027)
Talent
0.023***
(0.003)
0.019***
(0.003)
0.019***
(0.003)
0.019***
(0.003)
0.057***
(0.005)
0.056***
(0.005)
0.057***
(0.005)
Talent × Finance
Perceived stress
-0.004
(0.002)
Overwork more than 5 hours a week
0.016***
(0.002)
Overwork more than 10 hours a week
0.083***
(0.003)
Perceived job insecurity
Individual Controls
Year Fixed Effects
Observations
R2
(4)
-0.008*
(0.005)
Yes
Yes
80,406
0.700
Yes
Yes
80,406
0.706
Yes
Yes
80,406
0.713
Yes
Yes
80,406
0.707
This table reports the coefficient of standard OLS regressions. The dependent variable is the log of the
yearly gross wage. The oil, finance, chemistry, and consulting industries have a dummy variable. Talent
(which takes a value from 1 to 9) sorts schools based on level of recruitment, measured as the ranking
of the marginal student in the national competitive exam. All equations include year dummies, a female
dummy, a married dummy, a female × married dummy, a Paris area dummy, a working abroad dummy,
experience level squared and cubed, four hierarchic responsibility dummies, nine occupation dummies, four
firm size dummies, and four firm type dummies. Standard errors are in parentheses, * p<0.10, ** p<0.05,
*** p<0.01.
62
Table A6: . Engineering School List
School Name
Ecole Polytechnique
Mines Paristech
Centrale Paris
Ponts Paristech
Espci Paristech
Telecom Paristech
Supelec
Supaero (Isae) Toulouse
Institut D’Optique Graduate School
Ensta Paristech
Centrale Lyon
Centrale Lille
Ensae Paristech
Centrale Nantes
Centrale Marseille
Mines De Nancy
Mines De Saint-Etienne
Telecom Bretagne
Chimie Paristech
Ensica (Isae) Toulouse
Grenoble Inp - Ensimag
Agroparistech Grignon
Montpellier Sup Agro
Ensc Montpellier
Grenoble Inp - Phelma
Enac Toulouse
Grenoble Inp - Ense3
Grenoble Inp - PhelmaElectronique
Ensma Poitiers
Agrocampus Ouest
Enseeiht Toulouse Genie Electrique
Enseeiht Toulouse Electronique
Arts Et Metiers Paristech
Ensat Toulouse
Ensc Lille
Enscbp Bordeaux - Chimie-Physique
Ensea Cergy
Ensci Limoges
Ensi Poitiers Eau Et Genie
Insa Rennes
Insa Toulouse
Insa Strasbourg
Insa Lyon
Insa Rouen
Supmeca Paris
Ensc Rennes
Ensta Bretagne (Ex Ensieta)
Ensaia Nancy
Engees Strasbourg (Apprenti)
Ensiacet Toulouse Genie Industriel
Ensicaen Informatique
Enseirb-Matmeca Bordeaux Electronique
Ensic Nancy
Ensmm Besancon
Ensem Nancy
Enitab Bordeaux (Civil)
Eost Strasbourg
Cpe Lyon Electronique
63
Rank
Admission
Rate (%)
1
2
2
2
3
3
3
3
3
3
3
4
4
4
4
4
5
5
5
5
5
5
5
6
6
6
6
6
6
6
7
7
7
7
7
7
7
7
7
7
7
7
7
7
8
8
8
8
8
8
8
8
8
8
8
8
8
8
1.5
2.5
4.6
4.9
6.8
8.1
9.1
9.3
9.3
9.5
9.9
10.2
11.7
11.8
12.1
14.4
15.9
19.8
23.9
24.8
26.4
28.5
28.9
31.4
33.1
34.7
35.2
37.5
39.9
40.0
40.0
41.4
41.6
44.3
46.7
48.1
49.5
49.8
50.0
Baccalaureat
Grade (PostBac Schools)
16
16
16
16
16
50.0
52.4
52.6
53.2
53.5
53.8
53.9
54.1
56.2
56.8
57.6
58.0
58.3
58.4
Table A6 – continued from previous page
School Name
Ensp Strasbourg
Ensil Limoges Ee
Union Des Insa (2000)
Utc Compiegne
Inp Toulouse
Enesad Dijon
Isima Clermont-Ferrand
Ensiame Valenciennes Meca Energ.
Ensg Nancy
Ecpm Strasbourg
Ensc Mulhouse
Entpe Vaulx En Velin
Eivp Paris
Esial Nancy
Telecom St Etienne
Enssat Lannion
Telecom Sudparis - Cursus Evry
Vetagro Sup Clermont-Ferrand (Civil)
Agrosup Dijon
Ecole Des Mines Nantes
Estp Paris Topographie
Polytech Lille
Ecole Des Mines Douai
Ecole Des Mines D’Ales
Polytech Nantes
Ecole Des Mines D’Albi
Enstib Epinal
Polytech Paris-Upmc
Polytech Nice
Isat Nevers
Esil Marseille Biomedical
Epf Sceaux
Utt Troyes
Ensgsi Nancy
Estaca Levallois-Perret
Insa Val De Loire
Polytech’Montpellier
Ifma Clermont-Ferrand
Inpl Nancy
Ist Bretagne
Groupe
Polytech Marseille
Polytech Grenoble
Ensg Marne La Vallee
Eivl Blois
Enitiaa Nantes
Enise Saint-Etienne
Telecom Lille1
Ensccf Clermont-Ferrand
Eisti Cergy-Pontoise
Grenoble Inp
Polytech Orleans
Esstin Nancy
Esmisab Brest
Polytech Tours
Isty Versailles
Istil Epu Lyon 1
Polytech Clermont-Ferrand
Ensc Bordeaux
64
Rank
Admission
Rate (%)
8
8
8
8
8
8
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
10
10
10
10
10
10
10
10
59.1
59.4
Baccalaureat
Grade (PostBac Schools)
15
15
15
15
60.0
60.0
60.0
60.2
61.1
65.4
65.9
66.4
66.7
67.1
67.6
68.4
69.6
69.8
69.9
72.0
72.2
73.2
73.8
74.1
74.5
75.8
76.0
77.0
79.6
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
80.2
81.2
81.5
81.8
82.7
82.7
82.7
82.7
Table A6 – continued from previous page
School Name
Esirem Dijon Info-Elec.
Ensim Le Mans
Sup Galilee Villetaneuse
Lasalle Beauvais
Isen Brest
Isep Paris
Escom Compiegne
Hei
Ensisa Mulhouse Informatique Et Reseaux
Ensisa Mulhouse Textile Et Fibres
Eseo Angers
Ece Paris
Ensiie Evry
Eigsi
Ecam Lyon
Ensait Roubaix
Esigelec Rouen
Esme Sudria Ivry Sur Seine
Esb Nantes
Efrei Paris
Esiee Amiens
Esigetel Fontainebleau
Ei-Ispa Alencon
3Il Limoges
Grenoble Inp - Genie Industriel
Enit Tarbes
Polytech Paris Sud
Ecam Rennes
Enim Metz
Esilv La Defense
Ebi Cergy
Ensgti Pau
Esiea Paris
Itech Lyon
Itii Bass-Normandie Mecanique Ensicaen
Itii Picardie Mecanique Cnam
Itii Alsace Mecanique Insa Strasbourg
Itii Pays De Loire Inform. Ind. Eseo
Itii Pays De Loire
Enspm Rueil-Malmaison
Isa Lille
Polytech Savoie
Itii Alsace Informatique Loire
Isara Lyon
Ingenieurs 2000
Cefipa
Ifitep
Fiti2A Quimper
Itii Pays De Loire Btp
Cesi
Isupfere
Itii Aquitaine Prod. Maintenance
Isa Angers
Esgt Le Mans
Utbm Belfort-Montbelliard
Ist Vendee Mecanique Et Automatique
Esitapa Val De Reuil
Istp Ensme St Etienne
Esite Epinal
65
Rank
Admission
Rate (%)
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
82.7
82.7
82.7
83.1
84.4
85.6
85.8
85.9
86.4
86.6
87.1
87.9
88.0
88.9
89.8
92.6
94.5
95.0
96.0
96.6
97.0
97.6
98.2
98.5
100.0
Baccalaureat
Grade (PostBac Schools)
13
13
13
13
13
13
13
13
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
Table A6 – continued from previous page
School Name
Rank
Ist Toulouse
Cnam
Itii Champagne-Ardenne Mecanique Ensam
Itii Aquitaine Materiaux Enscpb
Itii Deux Savoies
Isel Le Havre
Itam
Isiv
Fip
Itii Bourgogne Genie Industriel
Dpe
Itiape Lille
Istimm
Ist Nord
Itii Lyon Informatique
Itii Aquitaine Mecanique
Itii Hte-Normandie Mecanique
Eia-Cesi
Igii Lens
Eme Ker Lann
Itii Centre Production Polytech’Orleans
66
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
Admission
Rate (%)
Baccalaureat
Grade (PostBac Schools)
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
Table A7: Distribution of Workers Across Industries
Firm Sector
1
2
3
Aero
Agriculture
Air
Alcohol
Auto
Carsale
Cement
Chemicals
Construction
Consulting
Drugs
Education
Elec Equip
Electricity
Electronic
Engineering
Finance
Food
Food Retail
Furniture
Glass
Health
Holding
Insurance
It
Machinery
Media
Metal
Mining
Misc. Bus. Services
Misc. Cons. Goods
Non Food Retail
Oil
Organization
Paper
Plastic
Printing
Public
Realestate
Restaurant
Ship & Rail
Soap
4.7
0.5
1.0
0.0
4.2
0.1
0.3
2.1
2.7
4.4
0.5
1.1
1.6
3.4
6.2
11.2
9.2
0.7
0.0
0.1
0.5
0.1
5.4
1.0
6.4
1.7
1.1
2.3
0.6
1.1
0.1
0.1
1.6
0.6
0.1
1.1
0.1
7.1
0.3
0.0
1.0
0.4
3.4
0.1
0.8
0.1
6.9
0.1
0.7
3.1
6.7
4.4
0.6
0.5
1.8
4.4
4.0
14.1
7.6
0.9
0.0
0.1
0.4
0.1
3.6
1.2
5.9
3.1
0.7
1.5
0.8
0.9
0.2
0.3
3.0
0.4
0.3
1.3
0.2
3.0
0.7
0.1
0.5
1.0
8.6
0.0
0.7
0.1
6.2
0.1
0.2
1.5
2.5
3.1
0.4
0.9
3.5
4.8
13.6
14.2
3.3
0.5
0.0
0.1
0.3
0.1
1.9
0.7
7.3
4.2
0.7
2.2
0.2
0.6
0.3
0.1
1.5
0.2
0.1
1.2
0.2
3.4
0.1
0.0
0.9
0.8
4
School Rank
5
6
7
8
9
10
Total
4.1
0.1
0.3
0.1
6.9
0.1
0.9
2.3
4.9
2.4
0.6
1.0
2.7
5.1
5.7
16.1
3.5
0.7
0.0
0.2
0.6
0.2
2.1
0.9
7.7
5.7
0.5
2.1
0.7
0.8
0.3
0.3
2.6
0.2
0.4
1.5
0.2
3.0
0.6
0.0
0.9
0.9
5.5
2.1
0.4
1.0
2.3
0.0
0.5
6.2
0.9
3.2
1.7
1.3
0.8
2.4
2.1
13.4
4.0
9.5
0.3
0.1
0.3
0.2
2.5
0.7
4.2
2.4
0.4
1.6
0.6
0.7
0.4
0.2
2.3
1.7
0.4
1.2
0.1
8.2
0.2
0.2
0.0
2.6
2.9
0.8
0.1
0.3
5.1
0.2
0.3
8.0
1.4
2.0
2.6
1.3
2.9
2.2
6.6
15.1
1.6
2.8
0.1
0.2
0.3
0.3
2.3
0.4
8.0
4.4
0.7
2.2
0.4
0.9
0.4
0.2
1.6
1.3
0.5
2.0
0.1
3.3
0.2
0.1
0.6
2.8
3.3
0.2
0.4
0.3
5.8
0.3
0.7
2.5
8.1
2.1
1.3
1.1
2.2
2.0
4.3
13.4
2.4
4.3
0.3
0.7
0.5
0.3
2.0
0.6
8.9
4.5
0.7
2.2
0.6
0.9
0.4
0.5
1.2
0.5
0.5
2.4
0.2
3.8
1.0
0.2
0.7
1.6
2.6
0.3
0.2
0.2
4.9
0.2
0.4
2.4
4.3
2.5
1.2
1.2
4.1
2.4
8.9
11.3
2.5
1.8
0.2
0.7
0.6
0.3
1.9
0.6
11.3
5.8
0.8
2.5
0.3
0.9
0.4
0.5
0.7
0.8
0.6
1.9
0.2
2.0
0.3
0.1
0.6
1.7
4.3
0.3
0.4
0.2
5.6
0.2
0.6
3.1
4.6
2.4
1.2
1.0
3.1
3.2
7.3
13.6
2.9
2.3
0.1
0.4
0.5
0.2
2.1
0.6
7.9
5.2
0.7
2.4
0.4
0.8
0.4
0.3
1.4
0.7
0.7
1.8
0.3
3.5
0.5
0.1
0.6
1.5
67
6.9
0.3
1.5
0.3
3.1
0.0
0.2
3.2
1.8
2.0
1.4
1.2
3.5
4.4
8.5
14.3
2.0
1.7
0.1
0.2
0.3
0.1
2.1
0.4
5.4
4.5
0.5
1.7
0.2
0.7
0.3
0.2
0.8
1.8
6.1
0.8
1.7
3.7
0.1
0.1
0.4
1.4
4.8
0.2
0.3
0.2
7.1
0.2
0.9
2.4
5.7
1.5
0.9
0.8
3.8
3.5
8.3
14.7
1.6
1.3
0.1
0.3
0.8
0.2
1.5
0.6
6.2
7.6
0.4
3.1
0.3
0.6
0.5
0.2
1.3
0.4
0.6
2.1
0.2
3.3
0.6
0.0
0.5
1.2
Table A7 – continued from previous page
School Rank
6
7
Firm Sector
(in % of workers)
1
2
3
4
5
Steel
Telecom
Textile
Transport
Utilities
Wholesale
Total
2.0
5.6
0.2
2.7
1.2
1.4
100.0
3.2
1.4
0.3
2.6
1.2
2.1
100.0
1.4
3.1
0.1
1.9
0.4
2.0
100.0
2.7
1.2
0.4
2.5
0.9
2.5
100.0
3.1
0.5
0.3
1.1
1.1
4.9
100.0
68
2.0
1.3
0.4
0.9
1.1
4.7
100.0
2.4
1.5
0.3
1.5
0.6
3.2
100.0
8
9
10
Total
1.4
1.5
0.3
0.8
1.7
4.9
100.0
1.9
1.8
0.6
1.6
0.7
3.6
100.0
1.5
3.0
1.4
1.3
0.5
5.0
100.0
2.0
2.0
0.6
1.6
0.8
3.7
100.0
Appendix E -
Model
This section develops a simple partial equilibrium model of the labor market in which firms compete for
talent. As in standard superstar models (Rosen (1981)), workers are heterogeneous in talent, and returns
to talent are increasing. This model differs in the assumption that the economy is composed of several
industries with heterogeneous output returns to talent. It develops predictions on the wage distribution
and wage elasticity to talent across industries.
Consider a one period economy composed of i = 1, ..., S industries. In each industry, firms produce
output by combining exactly one worker with talent T and capital. Specifically, the profit of a firm of
size k from hiring a worker of talent T in industry i is
π(k, T, i) = T i k αi − k − w(T, i)
where i ≥ 1, αi < 1. w(T, i) denotes the market wage of a worker of talent T in industry i. The unitary
cost of capital is one.
The key parameters in this model are i and αi . These values are determined by technology, and
vary across industries. αi represents talent scalability. If αi is high, an increase in one worker’s talent
will be matched by a large increase in the size of the firm he is working at. Talent scalability is higher in
industries with low physical constraints/replication costs. i represents the output elasticity of talent. In
industries in which the organizational structure is flatter, the impact of talent is higher and i is higher.
The assumption that i ≥ 1 is meant to capture increasing returns to talent, whereas αi < 1 implies
decreasing returns to scale.
All firms can observe the talent T of every worker in their industry. In each industry, at the beginning
of the period, firms enter freely at any firm size k. Next, all firms make simultaneous job offers and wage
bids to all workers in their industry. After that, workers decide which offer to accept.
The free entry of firms implies that a worker of talent T is hired by the firm of size k ∗ (T ) that is the
best match for his level of talent.
k ∗ (T, i) = arg max T i k αi − k − w(T )
k
which induces
1
1−αi
k ∗ (T, i) = αi
i
(5)
T 1−αi
Within each industry, competition among firms for workers’ talent ensures that firms earn zero profits.
The market wage of a worker of talent T in industry i is
w(T, i) = T i kT∗αi − kT∗
Introducing (5)
αi
1−αi
i
w(T, i) = T 1−αi αi
(1 − αi )
(6)
Consistent with Rosen (1981), wages are increasing and convex in talent T .
What does this process imply concerning the distribution of wages within each industries? One
measure of inequality is return to talent, which is given by
w0 (T, i)
=
w(T, i)
αi
1−αi
i αi
i
αi
1−αi
T 1−αi αi
i
T 1−αi
−1
(1 − αi )
i
w0 (T, i)
=
w(T, i)
T (1 − αi )
We can see that the greater talent scalability αi and output elasticity of talent i are, the higher wage
returns to talent are. If talent scalability αi is high, an increase in worker’s talent will be matched by a
large increase in the firm’s capital, which will increase the worker’s rent substantially. Similarly, as the
output elasticity of talent i increases, the higher the impact of talent on output. We also have
αi
i
∂2w
i
−2
1−αi
=
α
(
− 1)T 1−αi
i i
2
∂T
1 − αi
Prediction 1 As talent scalability αi or output elasticity of talent i increase, the more convex the wage
schedule is.
69
This prediction implies that when, in a specific industry, wage returns to talent are high, one should
observe more convexity in the distribution of wages.
The model also makes predictions in terms of inter-industry wage differentials. Let us consider two
workers with the same level of talent T working in two different industries, i and j, with i 6= j. The wage
ratio is given by
i
w(T, i)
− j
= T 1−αi 1−αj
w(T, j)
The wage ratio increases with the level of talent. Plugging the log, we obtain the central equation of the
paper:
j
i
ln(w(T, i)) − ln(w(T, j)) = (
−
)ln(T )
1 − αi
1 − αj
Prediction 2 For two workers of the same level of talent T working in industries i and j, with i 6= j,
j
w(T,i)
i
can be explained by the difference in wage elasticity to talent 1−α
− 1−α
.
the wage ratio w(T,j)
i
j
Predictions 1 and 2 have empirical implications concerning the distribution of wages within and
across industries. Let consider an industry in which wage elasticity to talent is relatively high (the ratio
i
1−αi is high) . Prediction 1 implies that the distribution of wages should be more skewed than in other
industries, Prediction 2 that the wage differentials should be higher at the top of the wage distribution
and that the wage premium can be explained by higher returns to talent.
70
Version du 4 mars 2009 20e enquête du CNISF sur la situation
socioéconomique des ingénieurs en 2008
Le questionnaire principal porte sur les données suivies régulièrement dans l’observatoire des ingénieurs (situation, emploi,
rémunération, satisfaction professionnelle)
Dans la mesure du possible les questions qui ne vous concernent pas sont occultées. Il vous faudra une quinzaine de minutes
pour y répondre.
Si vous souhaitez accorder un peu plus de temps à cette enquête (10 minutes supplémentaires), nous vous proposons
d’approfondir trois grands sujets d’actualité.
Souhaitez vous répondre aux chapitres optionnels suivants ? :
1. Compléments sur l’innovation (positionnement de votre entreprise et expérience personnelle)
1 - oui 2- non................................................................................................................................................. INNOV
2. Les critères d’attractivité des entreprises
1 - oui 2- non .......................................................... ATTRACT
3. Votre avis sur la situation économique
1 - oui 2- non ............................................................CRISE
I Signalétique personnelle
Votre année de naissance : 19 /__/__/ ............................................................................................................................ GEN
Vous êtes : .1 Un homme .2 Une femme .......................................................................................................................... SX
A quel pays correspond votre nationalité : /__/__/ ........................................................................................................ NAT
II Formation d'ingénieur
Votre formation à l'entrée en école d'ingénieur : ......................................................................................................... FOPR
1. Bac (prépas intégrées)
2. Classes préparatoires
3. DUT
4. BTS
5. Autres Bac+2 ou 3
6. Bac + 4 ou plus
7. Autre, dont diplôme étranger
Votre diplôme d'ingénieur a-t-il été obtenu : ..............................................................................................................FORM
1. En formation scolaire initiale
2. Sous statut d'apprenti
3. En formation continue
4. Par DPE ou VAE totale
Votre 1er diplôme d’ingénieur en France :
Sigle de l’école _______________________________________________________________________________ ASS1
Localisation (code du département) /__/__/ .................................................................................................................... DT1
Année de sortie /__/__/__/__/ .............................................................................................................................. PROMO1
Votre spécialité à l’issue de votre formation d’ingénieur : ............................................................................................ SPE
1. Agronomie, sciences de la vie, agro-alimentaire
2. Chimie, génie des procédés
3. STIC (Électronique, télécommunications, informatique, génie logiciel, mathématiques appliquées)
4. Électrotechnique, automatique, électricité
1 Version du 4 mars 2009 5.
6.
7.
8.
9.
10.
Génie civil, BTP, mines, géologie
Mécanique, production, productique
Physique, matériaux, énergétique
Économie, gestion, finance, audit
Généraliste, multiple ou sans spécialité dominante
Autre (bois, textile, eau, environnement…)
En quelle année avez-vous commencé à travailler comme ingénieur ? : /__/__/__/__/ ........................................... ANING
III Autre(s) diplôme(s) obtenu(s) après celui d’ingénieur
Commercial/gestion/management, type IAE, DESS, MBA, Master... : ............................................ 1 oui 2 non DIPGEST
Scientifique type DEA, DESS, Master : ..................................................................................................1 oui 2 non DIPSC
Mastère spécialisé de la Conférence des Grandes Ecoles :.............................................................................1 oui 2 non MS
Thèse ou PhD : ............................................................................................................................................... 1 oui 2 non TH
(Si TH=1) Cette thèse a t’elle été faite en France ......................................................................................1 oui 2 non THFR
Si vous avez obtenu un second diplôme d’ingénieur d’une école de France, précisez le sigle de l‘école .................... ASS2
Avez-vous aussi obtenu un double diplôme d’ingénieur ou un diplôme conjoint d’ingénieur à l’étranger ? 1 oui 2 non INGET
IV. Votre situation au 31.12.2008
Quelle était votre situation au 31-12-2008 ? ............................................................................................................... CSP
1 Activité professionnelle comme cadre
2 Activité professionnelle sans statut cadre
3 En recherche d'emploi
4 Sans activité professionnelle
1 Vous aviez une activité professionnelle comme cadre : ........................................................................................ CSP1
Salarié......................................................................................................................................................................................
1. Salarié en contrat à durée indéterminée
2. Salarié en contrat à durée déterminée
3. Préretraité en activité rémunérée
4. Retraité avec activité rémunérée
5. Titulaire de la fonction publique (Catégorie A)
6. Intérim, vacations ou contrat précaire
7. Contrat lié à une thèse : ATER, CIFRE, post doc….
8. Volontaire International en entreprise
9. Autre
Non Salarié
10. Travailleur indépendant, consultant, expert
11. Chef d’entreprise, gérant ou dirigeant
2 Vous aviez une activité professionnelle sans statut cadre : .................................................................................. CSP2
1. Agriculteur exploitant
2. Technicien, contremaître, agent de maîtrise
3. Ouvrier
4. Autre
3 Vous étiez demandeur d’emploi ............................................................................................................................. CSP3
4 Vous n’aviez pas d’activité professionnelle : ......................................................................................................... CSP4
1. En congé parental
2. Au foyer
3. Préretraité sans activité rémunérée
2 Version du 4 mars 2009 4. Retraité sans activité rémunérée
5. Invalidité, maladie longue durée
6. Autre
Avez-vous déjà interrompu votre activité professionnelle plus de 3 mois ? ........................................ : 1 oui 2 non FINAP
Si oui était-ce pour :
1. Maladie ou grossesse................................................................................................................................. RAISMAL
2. Chômage ................................................................................................................................................... RAISCHO
3. Vous occuper de vos enfants ou d'un membre de votre famille ................................................................ RAISENF
4. Formation ...................................................................................................................................................RAISFOR
5. Parce que votre conjoint changeait de lieu de travail ............................................................................. RAISCONJ
6. Autre .......................................................................................................................................................... RAISAUT
SI CSP =1 ou 2
Si vous ne travaillez pas à plein temps, indiquez pour quel pourcentage du temps plein /__/__/ %............................. TPS
En 2008, avez-vous fait des heures supplémentaires (rémunérées ou non) ................................................................ HSUP
1. Jamais ou rarement
2. Ponctuellement
3. Régulièrement, de 5 à 10h par semaine
4. Régulièrement, plus de 10h par semaine
V. L'entreprise qui vous employait au 31.12.2008 (Pour CSP =1 ou 2, uniquement)
Dans quel pays travaillez-vous (liste des pays) ...................................................................................................... PAYSW
Si c’est en France : indiquez le département) /__/__/................................................................................................... DTW
Nature de l'entreprise ............................................................................................................................................... NATEN
1. Entreprise individuelle ou travailleur indépendant
2. Secteur privé
3. Entreprises nationales (EDF, SNCF…), économie mixte, EPIC
4. État, collectivités territoriales, secteur public hospitalier
5. Autre, associations, ONG, organismes internationaux….
Secteur d’activité de cette entreprise au 31.12.2008....................................................................................................... SEC
1. Agriculture, sylviculture et pêche
Industrie
2. Énergie
3. Minerais, métallurgie, fonderie, travail des métaux
4. Production minéraux non métalliques, matériaux construction, céramique, verre
5. Industrie chimique
6. Industrie parachimique ........................................................................................................................................... SEC6
1. Dont parfum, cosmétique, arômes
2. Dont détergents, produits d’hygiène
3. Dont peintures, encres, vernis
4. Autre parachimie
7. Industrie pharmaceutique
8. Fabrication d'équipements mécaniques, de machines, d'armement
9. Matériel électrique, électronique, informatique ..................................................................................................... SEC9
S'agit-il d'industrie de défense ou de sécurité 1 - Oui 2 - Non
10. Constructions automobiles, navales, matériel de transport
11. Aérospatial ............................................................................................................................................................. SEC11
1. Aéronautique
2. Spatial
12. Industries agroalimentaires
13. Industries textiles, habillement, chaussures
3 Version du 4 mars 2009 14. Papier, carton, caoutchouc, matières plastiques
15. Autre
16. Bâtiment, travaux publics
Services et tertiaire
17. Grande distribution
18. Commerce, location de matériel, réparation, hôtellerie, restauration
19. Transports routiers, ferroviaires, aériens…
20. Télécommunications
21. SSII (Soc. de services et d'ingénierie en informatique) et éditeurs de logiciels
22. Ingénierie, sociétés de services aux entreprises autres qu'en informatique
Ce volet s’ouvre pour SEC=22 ................................................................................................................................... SEC22
Veuillez préciser votre activité propre au secteur de l’ingénierie (selon la convention SYNTEC)
1. Jeune ingénieur d’études, travaillant sous le contrôle d’un chef de projet
2. Ingénieur métier confirmé ou référant dans une discipline technique
3. Ingénieur projets impliquant un volume d’études inférieur à 50 000 heures
4. Chef de projet confirmé gérant des projets impliquant un volume d’études de plus 50 000 heures au moins
5. Directeur de projets gérant un ensemble de projets d’au moins 50 000 heures et rendant compte à la direction générale
6. autre
23. Assainissement, eau, gestion des déchets…
24. Assurances, banque, immobilier, holdings
25. Fonction publique d'Etat, territoriale ou hospitalière ............................................................................................ SEC25
1. Dont enseignement
2. Dont organismes de recherche
3. Dont autre fonction publique
26. Organismes internationaux
27. Autre tertiaire
Code NAF (ce code de 3 ou 4 chiffres et une lettre figure sur votre bulletin de salaire) : /__/__/__/__/__/ ..................NAF
Quel est le principal marché de cette entreprise ? ............................................................................................. MARCH
1 Le marché des biens ou services aux entreprises
2 Le marché des consommateurs finaux
3 Elle intervient sur ces deux types de marchés
Taille de l’entreprise (uniquement pour NATEN différent de 4) ............................................................................... TAILE
1. 0 à 19 salariés
2. 20 à 249 salariés
3. 250 à 499 salariés
4. 500 à 1 999 salariés
5. 2000 salariés et plus
L’établissement où vous travaillez est-il (uniquement pour NATEN différent de 1, 4 ou 5) : .......................... INDEP
1. Indépendant
2. Dépendant d’un groupe français
3. Dépendant d’un groupe étranger
4. Vous ne savez pas
Il y a 5 ans (fin 2003), dans quel secteur travailliez-vous ? : ................................................................................... SECW5
En quelle année avez-vous été recruté par cette entreprise ? : /__/__/__/__/ .............................................................. ANC1
Travaillez-vous dans la même entreprise depuis votre sortie de l’école ?...............................................1- Oui 2- Non UNE
Si non, dans combien d’entreprises avez-vous travaillé depuis votre diplôme : /__/__/ .......................................... NBENT
VI. Caractéristiques de votre emploi au 31-12-2008 (Si CSP = 1 sauf CSP1= 3ou 4)
4 Version du 4 mars 2009 Vos responsabilités
Vous avez le statut «cadre» .............................................................................................STATU1 oui 2 non 3 pas concerné
Vous avez des responsabilités de budget ? .................................................................. RESPCA1 oui 2 non 3 pas concerné
Vous avez des responsabilités à l’international ......................................................... RESPINT1 oui 2 non 3 pas concerné
Vous animez une équipe (sans responsabilités hiérarchiques) ...................................... RESPW1 oui 2 non 3 pas concerné
Vous êtes chef de projet .............................................................................................. CHFPRO1 oui 2 non 3 pas concerné
Vous prenez des décisions stratégiques ............................................................................ STRT1 oui 2 non 3 pas concerné
Vous êtes expert technique ................................................................................................. EXP1 oui 2 non 3 pas concerné
Vous êtes membre du comité de direction ou du directoire ............................................CODIR1 oui 2 non 3 pas concerné
Vous avez des responsabilités hiérarchiques ? ................................................................RESPH1 oui 2 non 3 pas concerné
et si oui ..........................................................................................................................................................................RESP
1. Vous encadrez une petite équipe
2. Vous encadrez un service ou un département
3. Vous avez des activités de direction générale
Combien de personnes encadrez-vous ? : /_/_/_/_/_/ ........................................................................................ NBENCA
Comment qualifiez-vous votre profil / carrière : .................................................................................................... CARR
1. D’expert technique
2. D’ingénieur polyvalent
3. De manager mobilisant une expertise technique importante
4. De manager mobilisant une expertise organisationnelle ou financière
5. Variable selon les époques
6. Autre type de carrière / de profil
Pensez-vous avoir des perspectives de promotion à court terme ? 1- Oui 2 Non 3 NSP .............................. PERSPRO
Si oui, ce sera ...................................................................................................................................................... PERSPRO1
1. En augmentant vos responsabilités dans votre spécialité d’expertise
2. En augmentant vos responsabilités dans votre activité dominante
3. En changeant d’activité et en allant vers :(ouvrir la boite de dialogue avec ACTDCOURT)
Production et activités connexes
Études, recherche et conception
Systèmes d'information, informatique
Commercial, Marketing
Administration, Gestion
Direction générale
Enseignement
Divers autres
4. En changeant d’entreprise
Votre activité ............................................................................................................................................................. ACTD
Quelle est votre activité dominante ? (Un seul choix)
1 Production et activités connexes
11 Production, exploitation, process, chantiers, travaux
12 Maintenance, entretien
13 Organisation, gestion de la production, pilotage, ordonnancement
14 Achats
15 Approvisionnements
16 Logistique
5 Version du 4 mars 2009 17 Qualité, hygiène, sécurité, environnement, développement durable
18 Autre
2 Études, recherche et conception
21 Recherche fondamentale
22 Conception, Recherche et développement
23 Ingénierie, études techniques, essais
24 Conseil, études non techniques, journaliste
25 Autre
3 Systèmes d’information (informatique et réseaux)
31 Production et Exploitation
32 Etudes développement et intégration
33 Support et assistance technique aux utilisateurs
34 Conseil en Systèmes d'Information, maîtrise d’ouvrage
35 Direction, administration et gestion
36 Autres
Pour ceux qui ont coché ACTD = 31 à 36
Dans quel domaine exercez-vous ces fonctions d’informaticien ? .......INFOSPE
1 Informatique industrielle
2 Informatique de gestion
3 Systèmes d'informations
4 Réseaux/ Télécommunications
5 Informatique embarquée
6 Internet, multimédia
4 Commercial, Marketing
41 Commercial, après vente, avant vente
42 Chargé d'affaires, chargé de marché
43 Technico-commercial
44 Marketing, communication produits
45 Autre
5 Administration, Gestion
51 Finances, gestion
52 Audit
53 Juridique, brevets
54 Communication d'entreprise
55 Ressources humaines et formation
56 Autre
6 Direction générale
7 Enseignement - formation
71 Enseignement supérieur
72 Autre enseignement
73 Formation continue
8 Autre
Quel est l’intitulé de votre poste : _________________
(merci de ne pas employer d’abréviations et de contrôler l’orthographe) ................................................................... POST
Cet emploi :................................................................................................................................................................. EMPL
1. Est votre premier emploi après votre formation
2. Fait suite à un recrutement externe
3. Fait suite à une mobilité interne
4. Est un emploi que vous avez créé
6 Version du 4 mars 2009 Depuis combien d’années êtes-vous dans ce poste ? /__/__/ .................................................................................ANACTD
Il y a 5 ans, fin 2003, quelle était votre activité dominante ? (selon la nomenclature de la question....... ACTD5) : /__/__/
(Pour ceux qui ont répondu 1 ou 2 à la question EMPL)
Si vous avez changé d'emploi dans les cinq dernières années, comment avez-vous trouvé l’information relative à cet
emploi ?............................................................................................................................................................... MODAL
1) Par candidature spontanée
2) Via un organisme (APEC, ANPE…)
3) Par un site spécifique d'emploi sur Internet
4) Par un site Internet d’entreprise
5) Par une annonce presse (passée par vous ou lue dans la presse)
6) Par votre école ou par les anciens élèves
7) Suite au stage ou à l’apprentissage dans l’entreprise
8) Vous avez été contacté par l’employeur ou un chasseur de têtes
9) Par un de vos proches (famille, ami)
10) Par une relation professionnelle
11) Suite à un concours
12) Autre (forum, salon…)
VII. Entreprenariat et Innovation (Pour CSP=1)
Entreprenariat
Etes-vous chef d'entreprise (salarié ou non) ? 1 oui 2 non ............................................................................... CHEFEN
Travailliez-vous dans une entreprise que vous avez
Reprise : 1 oui 2 non ............................................................................................................................................... CREERE
Créée : 1 oui 2 non ................................................................................................................................................. CREECR
Depuis combien d’années cette entreprise a t’elle été créée ou reprise ?......................................................... ANCREE
Dans les deux prochaines années, envisagez-vous de créer ou reprendre une entreprise ? .............................CREE2
1 oui 2 non. 3 Vous ne savez pas
Si vous envisagez de reprendre une entreprise, s’agit-il de celle où vous travaillez actuellement ? 1 oui 2 nonREPW
Avez-vous été préparé à créer ou reprendre une entreprise au cours de vos études d’ingénieur ? 1 oui 2 nonPREPCR
Innovation
Vous intervenez dans l’innovation et l’adaptation de process ? 1 oui 2 non .................................................................INNO
Vous intervenez dans la conception de nouveaux produits ? 1 oui 2 non ...................................................................CONC
Vous intervenez dans la conception de nouveaux services ? 1 oui 2 non................................................................... SERVI
Vous participez aux réflexions à moyen terme sur les nouveaux produits ou process ou services 1oui 2non.......... MTPRO
Votre activité vous conduit à travailler avec des designers 1 oui 2 non ................................................................. DESIGN
Votre entreprise emploie des salariés designers ou fait appel à des agences de design ?1 oui 2 non 3 nsp ............. AGDES
Votre entreprise a une politique de veille concurrentielle stratégique 1 oui 2 non 3nsp ....................................... POLVEIL
Vous participez à l’élaboration de sa stratégie technologique ? 1 oui 2 non ..............................................................STRTE
Dans les 5 dernières années (2004 à 2008) :
Vous avez été déposant ou co-déposant inventeur ou inventeur salarié d'un brevet ? 1 oui 2 non ...................... BREV
Vous avez participé à un programme de R&D de l'Union européenne ? 1 oui 2 non ..........................................PRUE
VIII. Rémunération 2008 pour l'emploi que vous venez de décrire (CSP=1 sauf CSP1 =3 ou 4)
Reportez le salaire cumulé brut pour l’année 2008 apparaissant sur votre bulletin de paie de décembre 2008 (salaire variable,
primes, 13e mois et avantages en nature inclus, hors intéressement et participation)
Salaire brut cumulé pour l’année 2008 : /__/__/__/ /__/__/__/ euros ................................................................... SALBR08
OU
7 Version du 4 mars 2009 Salaire net cumulé pour l'année 2008 /__/__/__/ /__/__/__/ euros ....................................................................SALNET08
SI ces montants ne correspondent pas à un temps plein ou à toute l’année 2008 :
À combien de mois de travail correspondent-ils ? /__/__/ mois (plafonner à 12) ......................................................... MS1
À quel pourcentage du temps plein ? /__/__/__/ % ....................................................................................................... PTP
Si vous avez une rémunération variable, quel pourcentage du montant annuel de votre salaire représente-t-elle ?
PVAR /__/__/__/ %
Est-elle déterminée ................................................................................................................................................ DETVAR
1. sur des objectifs strictement individuels
2. sur des objectifs collectifs
3. sur des objectifs à la fois individuels et collectifs
Quelle évolution de salaire envisagez-vous pour l'année 2009 ?................................................................................ EVOL
1. Stabilité
2. Une hausse de : /__/__/, /__/ % ..................................................................................................................... PHAUS
3. Une diminution de : /__/__/, /__/ ................................................................................................................... PBAIS
4. Vous ne savez pas
5.
Quels sont les avantages et primes dont vous avez bénéficié en 2008 ?
AVAN 1 à 16 1. Intéressement1 oui 2 non .............................................................. montant /__/__/__/ /__/__/__/ euros MONTAN1
2. Participation1 oui 2 non ................................................................ montant /__/__/__/ /__/__/__/ euros MONTAN2
3. Stock options 1 oui 2 non
4. Attribution gratuite d’actions 1 oui 2 non
5. Attribution d’actions à prix réduit 1oui 2 non
6. Retraite par capitalisation 1oui 2 non
7. Abondement du plan d'épargne d'entreprise1oui 2 non
8. Prévoyance santé 1oui 2 non
9. Voiture de fonction utilisable à titre personnel 1 oui 2 non
10. Logement 1 oui 2 non
11. Ordinateur portable1 oui 2 non
12. «Blackberry», iPod ou équivalent1 oui 2 non
13. Treizième mois (ou plus) 1oui 2 non
14. Prime exceptionnelle 1oui 2 non
15. Compte épargne temps 1oui 2 non
16. Autre 1oui 2 non
POUR LES RETRAITES (CSP1=4 ou CSP4=4)
Montant brut annuel de votre retraite ............................................................................ : /__/__/__/ /__/__/__/ euros RETR
Avez-vous une part de retraite par capitalisation.................................................................................... 1 oui .2 non CAPIT
Si oui, quel en est le montant annuel ......................................................................... : /__/__/__/ /__/__/__/ euros RETRC
POUR LES NON-SALARIES (CSP1 = 10 ou 11 )
Quelle est votre rémunération brute fiscale en 2008 : .......................... /__/__/__/ /__/__/__/ /__/__/__/ euros REMFISC
Si vous étiez salarié, quel salaire brut vous assurerait le même niveau de revenu
: /__/__/__/ /__/__/__/ /__/__/__/ euros SALEGU
IX. Vos satisfactions professionnelles (si CSP=1 sauf CSP1=3 ou 4)
SATIS 1 à 27
Parmi ces items, lesquels sont pour vous des sources :
Votre travail
1 de satisfaction 2 d’insatisfaction 3 indifférent ou sans objet
1. La sécurité de l'emploi
8 Version du 4 mars 2009 2. Le contenu du travail, l'intérêt des missions
3. Les opportunités de développement de votre carrière
4. Les missions à l'étranger
5. Les relations inter-personnelles
6. Le niveau de stress
7. La charge de travail
8. L'autonomie dont vous disposez
9. Les possibilités de faire évoluer vos compétences
10. La diversité des tâches à accomplir
11. La part de créativité de votre travail
12. L'exercice de responsabilités
13. La formation proposée par votre entreprise
14. Le sens, la valeur de votre travail
15. L’épanouissement personnel
16. Votre rémunération et ses compléments
17. L’équilibre travail/vie personnelle
L'organisation générale de l'entreprise
18. La pertinence de sa stratégie
19. La lisibilité de sa stratégie
20. La qualité de sa communication
21. La qualité de l'organisation générale de l'entreprise
22. Le style de management
23.
24.
25.
26.
27.
La façon dont vos propositions sont prises en compte
La reconnaissance de votre travail par la hiérarchie
La reconnaissance de votre travail par les autres ingénieurs
La reconnaissance de votre travail hors de l’entreprise (congrès, séminaires, activités d’enseignement…)
L’animation d’équipe et la gestion de projet
Diriez-vous que vous n'avez aucune cause d'insatisfaction majeure dans votre travail ?.................... SATNO : 1 oui 2 non
X. Mobilité et changement
Ne proposer MOB4ans que lorsqu’ANING est inférieur à 2004
Quelle(s) mobilité(s) avez-vous connue(s) ? .............................................................. MOB2008 1 à 12 - MOB4ans 1 à 12
1. Géographique, vers l'étranger.......................................................................... 1 En 2008 1 Entre 2004 et 2007 inclus
2. Géographique, dans le pays............................................................................. 1 En 2008 1 Entre 2004 et 2007 inclus
3. Vers un nouvel établissement (entreprise ou groupe) ..................................... 1 En 2008 1 Entre 2004 et 2007 inclus
4. Vers une nouvelle activité ............................................................................... 1 En 2008 1 Entre 2004 et 2007 inclus
5. Changement de niveau hiérarchique ............................................................... 1 En 2008 1 Entre 2004 et 2007 inclus
6. Vers un nouveau service ................................................................................. 1 En 2008 1 Entre 2004 et 2007 inclus
7. Vous avez perdu votre emploi (licenciement) ................................................ 1 En 2008 1 Entre 2004 et 2007 inclus
8. Vous avez retrouvé un emploi......................................................................... 1 En 2008 1 Entre 2004 et 2007 inclus
9. Vous avez changé d’employeur ...................................................................... 1 En 2008 1 Entre 2004 et 2007 inclus
10. Vous avez pris votre retraite ........................................................................... 1 En 2008 1 Entre 2004 et 2007 inclus
11. Vous êtes passé en préretraite ......................................................................... 1 En 2008 1 Entre 2004 et 2007 inclus
12. Aucun de ces changements ............................................................................ 1 En 2008 1 Entre 2004 et 2007 inclus
9 Version du 4 mars 2009 XI. Travail à l’étranger (si CSP=1 sauf CSP1 =3 ou 4 ET PAYSW NE1)
Si vous travaillez à l’étranger, êtes-vous parti à la demande de votre employeur ? ......................................... CHOIXETR
1. Oui
2. Non, c’est votre premier emploi
3. Non, vous avez quitté votre emploi pour aller travailler à l’étranger
4. Non, mais vous avez la possibilité de retrouver votre emploi en France (disponibilité, congé sabbatique…)
5. Non, vous étiez sans emploi quand vous êtes parti
6. Autre
Pensez-vous rechercher un travail en France dans l’avenir ? ....................................................................... RETOURETR
1. Non
2. Oui à long terme
3. Oui à moyen terme
4. Oui d’ici moins d’un an
5. Vous ne savez pas ou ceci ne dépend pas de vous
Comment estimez-vous votre situation à l’étranger comparée à celle que vous pourriez connaître en France :
En terme de
1Supérieure2 Equivalente 3 Inférieure ................................................................................. QUAL
1. Qualité de vie
2. Rémunération
3. Opportunités professionnelles
Avez-vous des inquiétudes pour
:
1oui 2 pas ou peu 3 pas concerné .................................................... PREOC1 à 4
1. Votre retour en France
2. La réinsertion professionnelle de votre conjoint(e)
3. L’éducation de vos enfants
4. Votre retraite
XII. Vos débuts dans la vie professionnelle (uniquement PROMO1 compris entre 2000 et 2008 inclus et pour FORM =1
ou 2)
Vous avez trouvé votre 1er emploi ............................................................................................................................. REMP
1. Avant votre sortie de l’école
2. Vous avez cherché........................................................................................................................ /__/__/ mois TEMP
3. Vous n’avez pas encore trouvé
En quelle année avez-vous commencé votre recherche :................................................................ /__/__/__/__/ ANRECH
Ce 1er emploi était dans l'entreprise où vous aviez fait un stage ou votre apprentissage ?...................... STAG 1 oui 2 non
Avez-vous trouvé votre premier emploi suite à une formation complémentaire à votre formation d'ingénieur ? MOYTROU
1 oui 2 non
Si votre 1er emploi n'était pas un emploi d'ingénieur avez-vous réussi à en obtenir un par la suite ? ...... SUIT 1 oui 2 non
Si oui combien de mois après avoir commencé à travailler ..................................................... : /__/__/__/ mois MOISING
Considérez vous que votre formation vous a suffisamment préparé à exercer l’activité que vous exercez ? ............ PREPA
1. Oui, totalement
2. Oui, plutôt bien
3. Suffisamment
4. Non, insuffisamment
5. Non, pas du tout
10