Chapter

2
Recruitment
“I sent the club a wire stating, ‘Please accept my resignation. I don't want to belong to any
club that will accept me as a member.’” (Groucho Marx)

Introduction

In this section of the book we are interested in how firms bring employees into the organi-
zation, and the patterns of careers they have once they are there. Table 2.1 presents
some data on these questions. The data are from confidential personnel records for all
management employees in a U.S. firm, over a 20-year period.1 Since the data are confi-
dential, we will refer to the firm as “Acme Incorporated.” We will present data from Acme in
several chapters to help illustrate concepts.

Number of
% who stay at Acme
years in
Hierarchical % of total % hired at
More
Level employees this level Current Only 1 Only 2 5-10
Acme than 10
position year years years
years
1 25.4 99.0 2.3 2.4 10.7 10.4 25.5 39.8
2 26.2 31.0 2.5 4.5 15.2 10.2 19.7 38.5
3 25.4 31.0 3.0 6.0 10.7 10.1 25.5 35.6
4 20.5 27.0 4.1 7.9 15.3 7.9 24.9 30.7
5-8 2.5 19.0 4.0 9.7 7.1 14.3 42.9 28.6

Career Patterns at Acme, Incorporated

Table 2.1

Acme is in a service industry. Its management ranks have 8 hierarchical levels, from entry
(Level 1) to CEO (Level 8). Most management employees are in the first four levels. Level
1 is what is often called a port of entry; virtually all who work in jobs at this level were hired
into Acme at this level. Since it is the bottom of the management job ladder, this should not
be surprising, as demotions are very rare in most companies. At Levels 2-8, most em-
ployees were not hired from the outside, but instead were promoted from within.

Because of promotion from within, managers in upper levels have substantial experience
at Acme on average. For example, Level 4 managers have almost eight years of expe-
rience in the company. It also appears that movement between levels (promotion) is more

1
The Acme tables are based on Baker, Gibbs & Holmstrom (1994a,b). The Acme data are from a single firm, but
the patterns that we illustrate with the Acme data in this book appear to be fairly representative of the policies in
many firms, in different countries. See the papers cited in Gibbs & Hendricks (2004).
 rapid at lower levels, since the average number of years in the current job is longer in up-
per levels.

The last four columns give some sense of turnover and career length for Acme’s manag-
ers. There are two patterns. First, many leave Acme very quickly after being hired. For ex-
ample, about 11% of those who are hired at Level 1 leave within the first year, while
another 10% leave the next year. Conversely, if a manager stays at Acme beyond the first
year or so, he or she has a good chance of staying for many years. For example, about a
quarter will stay for between five and ten years, while about a third will stay for more than
ten years.

Thus, there seems to be evidence of sorting in the first few years on the job. Almost a
quarter of new hires will leave within two years, either because Acme decided it did not
want to keep them, or because new employees decided they did not want to stay. Second,
if employees survive the sort, they often enjoy careers at the company that last for many
years. This suggests there is some value to having employees remain with the firm. These
are issues we explore in this and the next three chapters.

With this introduction, let us now return to the issues we raised in Chapter 1. Once your
firm has decided which types of workers it would like to hire, it must recruit for those types.
There are two general issues. First, how can it weed out undesirable applicants? In some
jobs, hiring the wrong type of employees can cause major problems, disrupting output and
costing the firm not only wages but lost profits. Second, how can the firm attract the right
types of applicants? Attracting the right type will reduce workforce problems, as well as re-
cruitment and turnover costs. Put another way, your firm must sort new hires, just as Acme
apparently does. How can you think about sorting for the most effective workforce?

Screening Job Applicants

One strategy for attracting good quality job applicants is to offer a high level of pay or ben-
efits. This will generate a large pool of applicants, and higher quality applicants will be
more likely to apply for the job than if pay was lower. Unfortunately, so will low quality ap-
plicants. The personnel office will be flooded with resumes, and only a small portion may
be qualified. Some undesirable workers will slip through the hiring process and become
employees, while some desirable workers will get lost in the shuffle and never hired. This
is not a very useful approach just by itself.

The problem of the wrong type of applicant applying to the firm is called adverse selec-
tion.2 This is a general problem in economics, not just in employment. The problem arises
because of asymmetric information. One party knows what type they are (in this case, a
high or low quality job candidate), and the other does not. The one that knows uses this in-
formation strategically to personal advantage. A classic example involves used car sales.
Owners know the quality of their used car. Those with good quality used cars are more
likely to keep them, while those with low quality used cars are more likely to sell them. This
implies that the quality of used cars is lower than it would otherwise be. It also implies that
owners of high quality used cars may find it difficult to get a good price for their car, since
buyers worry that it is a low quality car.

2
In 2001, George Akerlof was awarded the Nobel Prize in Economics for analyzing the problem of adverse selec-
tion. He shared it with Michael Spence, who was awarded the prize for analyzing the problem of signaling that we
discuss later in this chapter. Joseph Stiglitz also shared the prize that year, for analyzing problems of asymmetric
information.

23
 Adverse selection arises in our case when the wrong kinds of workers are attracted to the
firm. A number of approaches can be used to mitigate the problem of adverse selection in
recruiting. Let us begin with a simple case first, the use of credentials.

Credentials

An obvious approach to weeding through resumes of job applicants is to look for creden-
tials that distinguish some applicants from others. The most important ones are generally
the type of experience (job and promotion history) the applicant has, the type of training
(e.g., college major or MBA), and the quality of school the applicant attended. Indeed,
these are almost always the most important lines on anyone’s resume. What makes a
credential useful for hiring? Here are some considerations:

Informativeness of the Credential

Ability to perform well on the job must be positively correlated with ability to obtain the cre-
dential. For example, a university degree is a useful credential only if university graduates
tend to be more productive at the job in question. A credential can be informative in two
ways. First, it may mean that the holder of the credential has knowledge or skills that apply
directly to a job. This might be the case with a CPA or MBA. Second, it may mean that the
holder of the credential has innate abilities that tend to make one more productive on the
job. An example might be a high score on an aptitude test, or winning a scholarship.

Cost of Obtaining the Credential

A valuable property of the credential is that it is relatively easy for well-qualified workers to
obtain, compared to poorly qualified workers. When this is true, the credential is very likely
to signal differences in ability. For example, it is not very difficult for a qualified accountant
to pass the CPA (Certified Public Accountant) exam, but it is virtually impossible for some-
one with no training in accounting to pass. Thus, using the CPA as a screen effectively
sorts between qualified and unqualified accountants.

On the other hand, a credential that is extremely expensive for all workers to obtain will not
do well in sorting them. If a credential is very difficult to obtain, few applicants will have it.
For a credential to be effective, it must be that most qualified applicants possess the cre-
dential while most unqualified ones do not. If a very small subset of qualified applicants
has the credential, or a large subset of unqualified ones does, the credential is not helpful.

Return on investment in the credential. If the difference in wages between those who have
credentials and those who do not is not very great, small differences in credentials will sig-
nal large differences in ability. For example, if the credential is education, and the increase
in earnings from obtaining a college degree is small, only the most talented will get the de-
gree. This is because they are the ones for which obtaining the degree is cheapest. When
the rewards to obtaining a degree are large, even not-so-able people can be induced to
get the degree.

We will see these ideas below when we discuss signaling. Signaling is one way of resolv-
ing adverse selection problems. In many cases, a high quality type can incur some costs
to signal to others that they are high quality. If low quality types do not also invest in the
signal, then it can serve to distinguish the high quality types from the low quality types. For
example, an owner of a high quality used car can offer a warranty. It may also be possible
for a high quality job applicant to signal this to potential employers. Before we discuss this
idea, it is useful to consider simpler sorting issues first.

24
Learning a Worker’s Productivity

Suppose that you advertised an investment banker job as in Chapter 1. In response you
received a set of resumes from job applicants. You looked through them, and selected a
subset with appropriate credentials. In a job like investment banking, small differences in
ability, personality, or other employee characteristics may lead to large differences in effec-
tiveness on the job. Unfortunately, the self-selection of applicants to the job, and further
winnowing by sifting through their resumes, makes the pool of remaining applicants look
more and more alike. In general, the more that a set of applicants has already been
sorted, the lower will be the variance between the remaining candidates. What should you
do next?

You could hire one at random, and take a chance. However, given the stakes it is likely to
make sense to expend some resources to screen them further.

There are a variety of methods that firms use to screen applicants. Some give job appli-
cants tests, to see how they perform on specific tasks. This approach is more likely to
work well for jobs with fixed, measurable tasks. It is not likely to work well for an invest-
ment banker. Many firms use psychological profiling. Unfortunately, this technique does
not tend to work well in practice. One reason is that psychology is a highly inexact science.
Another is that job applicants have an incentive to game the test, trying to sound like better
employees than they are in reality. For example, one study found that 90% of job appli-
cants who took one popular psychological test were able to inflate their “conscientious”
score.3 Finally, virtually all firms conduct personal interviews of job applicants. Such inter-
views can vary from simple to elaborate. In the case of investment banking, applicants
may be put through several rounds of interviews, eventually being flown to the company’s
headquarters to meet with high-level partners over several days. Such a process can be
extremely expensive.

All of these examples involve some costs (except when the firm hires applicants without
any screening whatsoever). Consider the following example, and think about the extent to
which your firm should invest resources in screening applicants carefully.

Screening Bankers

Table 2.2 shows productivity levels for five hypothetical types of job applicants (A through
E) in two different firms, an investment bank and a commercial bank. Assume that the re-
maining applicants (after earlier rounds of weeding out) are paid about £100,000, so each
bank expects to pay about the same salary to anyone it hires.

Type
A B C D E
% of Job Applicants 10% 20% 40% 20% 10%
Productivity Investment Bank –£250 0 125 200 450
(£ thousands) Commercial Bank £95 100 110 120 125

Screening Investment Banking Job Applicants

Table 2.2

3
See Paul (2004).

25
Finding out what type a job applicant is has obvious value to each firm. The investment
bank would want to avoid A and B types, because productivity would be lower than pay,
while the commercial bank would want to avoid A types.4 Suppose that applicants can be
put through a series of tests that cost £2,000 per person, and give definitive information on
which type the applicant is. How valuable is such information? In other words, how much
would each be willing to pay to screen workers before hiring? Table 2.3 provides figures to
help us answer this (all numbers are expected values rounded to the nearest £100).

Screening
Screen? Productivity Salary Profit
Cost
Investment No £110 £100 £0.0 £10.0
Bank Yes 193 100 2.9 90.1
Commercial No 110 100 0.0 10.0
Bank Yes 112 100 2.2 9.8

Profitability of Screening at an Investment Bank & a Commercial Bank

Table 2.3

With no screening, both banks would have £110,000 average productivity from each new
hire, or average profit of £10,000.

With screening, the investment bank would reject types A and B, and accept only 70% of
all applicants. The average productivity of the C, D and E types hired would be about
£193,000, substantially higher than without screening. The screening cost per worker ac-
tually hired would be £2,000·10/7 (since the bank would hire an average of 7 out of every
10 applicants), or £2,857 per hire. Average profit from each new hire would rise with
screening to about £90,100. The investment bank would profit greatly from screening ap-
plicants.

If the commercial bank screened applicants, it would reject those of type A, hiring 90% of
all applicants. Average productivity would rise only slightly to about £112,000, but at a
screening cost of £2,000·10/9, approximately £2,222, per new hire. Profit per new hire, net
of screening costs, would fall to about £9,800. The commercial bank would not benefit
from screening.

Why the difference? There are two reasons. First, the investment bank wanted to screen
out three times as many workers as the commercial bank. The point of screening is to
avoid hiring the applicants who would not be profitable. Second, the downside from hiring
poor candidates was worse at the investment bank; some applicants would have pro-
duced nothing, and others would have destroyed value. The investment bank was more at
risk from hiring the wrong type of worker.

This example motivates issues to consider when screening (see the Appendix for a formal
treatment):

Screening is more profitable when the test is more effective: A test can be more effective
in several ways. First, it can be cheaper to administer. Second, it can be more accurate.
That is, it can correctly distinguish between desirable and undesirable job applicants a

4
You should remember from Chapter 1 that each bank would want to hire those with the highest productivity per £
of compensation cost. That is true; the firm will also want to keep hiring as long as expected productivity is greater
than expected employee costs (profit from the hire is positive).

26
 higher percentage of the time. No test is 100% accurate. Moreover, as noted earlier, job
applicants often try to game such tests so as to appear to be a better candidate than they
really are. Finally, an effective test is more discriminating. That is, it weeds out a higher
fraction of candidates, recommending a smaller fraction for hiring. In the example above,
the commercial bank’s screen was not very valuable because only 10% of candidates
were rejected.

Screening is more profitable when the stakes are higher: The purpose of screening is to
avoid the unprofitable candidates. Therefore, the greater the downside risk from hiring the
wrong person, the more value there is to screening. Similarly, the longer that a new candi-
date can be expected to stay with the employer, the more valuable will be the screen.
Firms that intend to hire employees for the long term thus tend to invest more in careful
screening before committing to a new hire.

Is Screening Profitable? For Whom?

If the investment bank screens workers, productivity is much higher for its employees than
for random job applicants. Now the same issue arises that we faced with our risky hire in
the previous chapter, Svensen, when she turned out to be a star. The labor market will
value our screened employees more highly, simply because we decided to hire them (they
passed our screen). Therefore, it is not realistic to assume that we can continue to pay
£100,000 at the investment bank, if productivity is almost double that. Other investment
banks would bid away our workers, once they realized that we screen our employees
carefully.

What will we end up having to pay our employees? It is hard to tell without additional in-
formation. It is even conceivable that we would have to pay as much as their productivity,
£193,000, if the labor market is very competitive. Screening may not always be profitable
to the employer; indeed some firms screen extensively, while others hardly screen at all.

What about job applicants? Why would they apply to a firm if they knew they would be
screened? It must be that the potential higher pay, if they pass the screen, compensates
for the trouble and risk involved in the screening. If the application process is not too diffi-
cult, then the extra compensation need not be too high to make it worthwhile for applicants
to try. If the screen is extensive, such as probation (described below), however, job appli-
cants may have to be compensated significantly in order to be willing to undergo the
screening process.

When the firm cannot benefit much from the screen because labor markets are competi-
tive, job applicants will have to pay for much or all of the screening. Of course, this hap-
pens already in the case of pre-job market screening such as education or professional
certification. But it may also happen with on-the-job screening. Workers can pay for the
screening implicitly, by their willingness to accept lower pay during the screening period
then they would otherwise earn.

In any case, it is likely that both the employer and its employees will share the benefits
(and costs) of screening. Firms that screen more extensively will tend to pay more, both
because their employees are more productive, and because applicants will require some
compensation for the costs and risks of trying to successfully earn long-term employment
at the firm.

A further consideration arises when employees have some idea about whether they are
high or low ability. Those who are high ability have a better chance of passing the screen,

27
 so they have more to gain from the screening. Thus, they should be more willing to under-
go and pay for the screening. We discuss this below under the topic of signaling.

Probation

The screening methods described above may be useful, but are imperfect. An important
concern is that they are only proxies for what the firm really cares about – how the person
actually performs the job. In many cases, the only way to truly tell if a job applicant is a
good fit for the job is to have them perform the job itself. Thus, a final approach to screen-
ing is to have the job applicant do the job for some period, either very briefly during inter-
views, or more extensively during some testing period. The most elaborate form of this is
to hire a worker for a probationary period, and only hire them long-term if their perfor-
mance during probation is adequate.

Of course, a problem with probation is that costs of terminating employees can be sub-
stantial. In Italy, if the firm is found to have fired an employee who has worked for the firm
long enough without legal cause, the firm must rehire the worker, pay lost wages and so-
cial insurance contributions, and pay a penalty to the government. In Indonesia, firms must
pay workers severance of more than one month of salary for each year that the employee
worked for the firm, up to 9 months of severance, plus 15% of salary as “worker’s rights
replacement money.” By contrast, turnover costs are usually very small in Denmark. The
general trend has been toward increased turnover costs, due to greater employment regu-
lation and litigation over wrongful termination.

Where firing costs are high, employers can often still use probation, under a different form.
For example, employees might be hired on a temporary basis through a temporary em-
ployment agency. Those who perform well could then be offered regular employment.
Those who do not perform well do not need to be fired; they are just not hired from the
temporary agency. Indeed, some temp agencies have an explicit strategy of serving as a
screening agency of this kind for employers.

A similar approach that firms can sometimes use is to hire applicants on temporary (fixed
term) contracts. When the contract term is over, the firm can elect to hire the worker more
permanently, offer a new fixed term contract, or not rehire the worker at all. Such contracts
are not just used for low skilled jobs; many firms hire high-level consultants in a similar fa-
shion.

Evidence suggests that increasing regulation of employment is one factor behind the re-
cent growth in temp agencies worldwide. For example, in the United States some em-
ployment regulations vary from state-to-state. Those states that regulate termination more
strongly tend to be the ones where firms are more likely to use temporary workers. In Eu-
rope, the employment relationship is more highly regulated than in most other parts of the
world, and hiring of temps is quite common. According to one study, 13 percent of all wage
earners in the European Union were employed on temporary contracts. In Spain, the cor-
responding figure was 31%, and roughly half of workers under the age of 30 were em-
ployed on temporary contracts.

28
 Reducing Firing Costs in France

In September 2005, the French government passed a new law designed to make it
easier for companies with 20 or fewer employees to hire and fire workers. It speci-
fied a “New Recruit Contract” allowing such companies to lay off workers anytime in
their first two years in the job, for any reason. Those laid off were to be given at least
two weeks’ notice, and would be entitled to unemployment benefits, but would not
have to be given the level of severance pay that is standard for other French work-
ers.

Labor unions and opposition party leaders criticized the law, which was initially
passed by decree under a new “emergency procedure” that allows the French gov-
ernment to enact employment legislation without consulting lawmakers. In April
2006, students, union members, and others marched in protest on the streets of
Paris. In response the law was cancelled by President Jacques Chirac.

Source: Associated Press (2005-6)

Implications of Probationary Screening

If the firm uses probation to screen workers, and retains those who are good performers,
there are several interesting implications. First, the firm is likely to promote those who sur-
vive the screening. They have been revealed to be more productive than average job ap-
plicants. Once this is evident, the firm is likely to give them greater responsibility.

Second, the system will generally be up-or-out, since those not promoted will typically not
be rehired. This is much like the lower levels of promotion ladders that are seen in most
professional service firms.

Third, a large raise in compensation will usually accompany the promotion. The firm pro-
motes those it finds most productive, so a promotion implies that you are more talented
than the average new hire, which raises your market value. For this reason, firms typically
have to offer raises on promotion, or risk losing those promoted. Moreover, since promo-
tion is based on performance, and performance depends in part on an employee’s effort
on the job, the promotion will become a form of incentive pay. We return to these issues in
Chapter 11.

Signaling

Most people have a good idea about their skills, work ethic, and ambition – the things that
make them a good employee. Let us now assume that workers know what type of em-
ployee they will be. If workers know what type they are, and share this information honest-
ly with employers, a firm could recruit employees of a certain type by simply putting up a
notice that it is looking for workers of that type. Unfortunately, this is not likely to be an ef-
fective approach. Recall our discussion earlier in this chapter about offering high wages to
high skilled workers. A firm that tried this would find itself facing adverse selection, since
job applicants who were not highly skilled would be tempted to apply anyway. This is why
we decided that some kind of screening would be necessary.

29
When workers know more about their employability than employers do, screening can be
used to solve this adverse selection problem. After all, screening works by sorting workers,
and keeping those who fit well and are most productive with the firm. It then pays higher
wages to those who are screened out. Shouldn’t this attract those who are good candi-
dates to apply in the first place, and deter those who are not good candidates? Let us
consider an example to see how this might work.

Consider a simplified example of recruiting for an investment bank. Suppose that simple
interviewing allows the bank to weed out Types A through C easily, but that it is much
harder to distinguish between Types D and E without further screening. The bank would
like to hire E types because they are the most profitable. Instead of screening, can it con-
struct a job offer involving probation, up-or-out promotion, and a raise on promotion that is
attractive to the E types, but not to the D types?

To model this situation, we need a little more information. First, let us assume that the
bank can figure out what type an employee is after observing them on the job for one year.
However, the accuracy of this judgment is not perfect: 10% of the time, the wrong decision
is made. Thus, 10% of D types are promoted when they should not be (does this remind
you of your boss?), and 10% of E types are not.

We also need to know what each type could earn elsewhere, since we have to offer a
more desirable package to E types, but a less desirable package to D types. Assume that
D types can earn £175,000 in other jobs, while E types can earn £200,000 in other jobs.
Thus, their alternatives working elsewhere for two periods are twice these salaries,
£350,000 for D types, and £400,000 for E types.

Finally, we need to know how long those promoted will work for us. To keep things simple,
assume that they will work for one year after promotion. Table 2.4 provides these figures
(rounded to the nearest £1000), and calculates the expected value of the job offer to each
type, for different salaries in the two years W1 and W2.

Type
W1 W2 D E
Expected pay Expected pay
Apply? Apply?
Alternative Apply Alternative Apply
£200 £200 £350 £378 yes £400 £400 no
180 225 350 360 yes 400 403 yes
160 250 350 343 no 400 405 yes
140 275 350 325 no 400 408 yes
120 300 350 308 no 400 410 yes
100 325 350 290 no 400 413 yes

Motivating Self Selection of Job Applicants

Table 2.4

The first offer considered equals what E type applicants could earn elsewhere, £200,000
per year. This obviously attracts D types, but is no lure to E types. The second offer lowers
pay during the probation period (W1), and raises pay after promotion (W2). Each row below
further lowers W1 and increases W2. Because the promotion is not guaranteed, the firm
must offer more than £400,000 in total pay W1 + W2 in order to attract E type applicants.
For this reason, and reflecting the risk involved in accepting lower initial pay compared to

30
what E types could earn elsewhere, each subsequent row involves higher total pay for
those who win promotion.

To calculate the actual values to D and E if they apply, note that each earns W1 in period 1.
In period 2, D types earn W2 with 10% odds, and their alternative wage with 90% odds.
Similarly, in period 2, E types earn W2 with 90% odds, and their alternative wage with 10%
odds:

Value of applying for D = W1 + 0.9·£175,000 + 0.1·W2;

Value of applying for E = W1 + 0.1·£200,000 + 0.9·W2.

The first two offers are attractive to D types. This is because pay is higher at the invest-
ment bank than they could get elsewhere, even during probation. Our first lesson is that in
order to deter some applicants from applying, we must pay less than those applicants can
get elsewhere before probation.

Similarly, the last few offers are attractive to E types, because the high pay after promotion
is enough to compensate for the low initial pay, given the high probability that E types will
be promoted. Our second lesson is that in order to attract some applicants, we must offer
more than they could earn elsewhere after probation.

Thus, probation can generate good self selection of job applicants, thereby solving the ad-
verse selection problem, if we pay a sufficiently low amount during probation, and a suffi-
ciently high amount after probation. One way to understand this is to note that in effect the
firm is demanding that each applicant post a bond – by accepting less than they could
earn elsewhere – during probation. In return, if they perform well and are promoted, the
firm will give them a reward – by paying them more than they could earn elsewhere. Fig-
ure 2.1 shows the type of contract that we are considering.

W2 > W E

Deferred reward

WE
Outside
Up-front bond alternatives
WD
W1 < W D

Probation Promotion

NPV of job depends on probability of getting

promotion, which differs by employee type

Deferred Pay as a Screening Mechanism

Figure 2.1

31
 Note in Figure 2.1 that the higher-ability employees of type E receive a smaller reward on
promotion, and pay a larger cost during probation, than do lower-ability employees of type
D. The up-front bond W – W1 is larger for E types, since their outside alternative is larger.
Similarly, the deferred reward W2 – W is smaller for E types for the same reason. If the D
types would pay a smaller bond, and earn a larger reward if promoted, than how can this
type of job offer deter them from applying, while motivating E types to apply? The answer
is that the firm must put the employees through a rigorous performance evaluation before
awarding promotion. The evaluation must result in sufficiently high probability that E types
will be promoted, and sufficiently low probability that D types must be promoted. The low
success rate for D types reduces the expected value of the job for them compared to E
types.

This discussion illustrates the general economic idea of signaling. Signaling is a method
that can sometimes be used to solve adverse selection problems. The high quality type
signals his or her type to the market by incurring a cost. If low quality types are not willing
to incur this same cost, then the signaling is effective: the fact that someone incurs the
cost proves that they are the high quality type.

Who Pays, and Who Benefits?

Signaling only works if the incentives of job applicants are addressed: Type D’s must be
deterred, but Type E’s must be motivated to apply. Thus, it is the employee who will pay
for most or all of the cost of signaling, and enjoy most or all of the benefit of the signal. In
Figure 2.1, we see that employees pay for the signal by accepting wages below what they
could get elsewhere initially. They also pay in the sense that they incur some risk that they
will not get promoted, even if they are high ability (if the screening is imperfect). They are
rewarded later by earning more than they could earn elsewhere after promotion.

The employer may also pay for part of the probation, and get some of the benefits.
Whether this happens depends on what average wages end up being compared to prod-
uctivity. The firm benefits during the probation period from pay that is below productivity,
but incurs costs when it pays more than productivity to those that are promoted.

Examples

An example of signaling is the seller of a used car who offers a warranty. The warranty is
costly to the seller, and is the signal. The fact that the seller is willing to offer the warranty,
while some other sellers are not, may signal that the car is of higher than average quality.
In our employment example, Type E’s can signal their type (and confidence in their ability
to perform well and earn promotion) by their willingness to accept low pay in the first pe-
riod. This only works if the D types are not also willing to accept the same contract.

There are many applications of signaling in the business world. For example, venture capi-
talists typically demand that entrepreneurs invest all of their family’s personal funds in a
new business venture. They may even demand that the entrepreneur mortgage his or her
home, and invest the proceeds in the startup. At first glance this may seem odd – isn’t the
venture capitalist supposed to provide the funds? Demanding that the entrepreneur put
some “skin in the game” is important, however, as it helps the venture capitalist separate
out the most confident and serious candidates from the least.

Another example involves joint ventures between two firms. In such cases, it is common
for both firms to invest some funds. One reason for this may be so that each firm can sig-
nal to the other the seriousness of their intentions to make the joint venture profitable.

32
 As discussed above, probation is a form of pay for performance. It can serve as a signal if
high ability job applicants are willing to accept a job offer with a risky but potentially lucra-
tive promotion ladder. The same idea applies to any form of pay for performance. If a new
hire is willing to accept more risky pay for performance, this may signal that they are high-
er ability, and believe themselves to be a good fit for the job. The opposite is true if they try
to renegotiate to reduce the pay for performance. Thus, pay for performance helps sort
employees in addition to motivating them.

Education as a Signal

Education is an important potential example of signaling. Assume for the sake of argument
that students learn nothing useful in school. However, suppose that more talented stu-
dents find it easier to learn the material quickly. If so, then they might be able to signal their
talents to the labor market by investing in more education than less talented students. In
this view of education, instructors require students to pass through increasingly difficult
screens. At each stage, some find the cost of passing the next screen to be too high, so
they do not get that level of education. Those who find the screen relatively cheap do
enroll for that level of schooling. The labor market recognizes this, paying more to those
who have obtained more schooling.

Indeed, as Table 1.3 in the last chapter showed, those with more schooling do earn more.
Is this due to signaling? It is possible, but it seems extremely implausible that this is the
only explanation. If the only purpose of education is to screen, we could probably find
more efficient ways to do so than to have students go to college for four years, for exam-
ple: we could give them a large test at the end of high school. In Chapter 3, we consider
investments in new skills. Education clearly plays an important role in doing this.

That said, there is evidence that education does have some role in screening workers. For
example, those who almost complete four years of college earn a little less than those who
go a little bit further and complete their degree – there is a discrete jump in earnings asso-
ciated with earning the formal credential. This is hard to explain by training alone.

Signaling More Formally: Separating and Pooling Equilibria

Let us consider a formal example of signaling so that we can see how it works. Suppose
that junior accountants can invest in some education or on-the-job training. If they com-
plete the training, they become Certified Public Accountants (CPAs). Assume that there
are only two types of accountants, “quicks” and “slows,” depending on their ability as ac-
countants: quicks are more productive, and also find it easier to obtain the training neces-
sary to pass the CPA exam.

To formalize these ideas, denote the present value of the employee’s productivity as Q,
and the cost of obtaining the CPA credential as C. Subscripts refer to the two different
types of accountants.

Assume that the labor market pays accountants a salary that exactly equals their expected
productivity. The fraction of accountants who are quicks is α, so the fraction who are slows
is 1–α. Thus, if the labor market cannot tell the two types of accountants apart (if there is
no signaling), pay equals:

Average productivity = Q = α·Qq + (1–α)·Qs.

33
On the other hand, if the quicks do succeed in distinguishing themselves, they will be paid
their productivity Qq. Those who do not signal will then be assumed to be slows, so any-
one who does not signal is paid Qs.

The quicks would like to distinguish themselves from the slows, so that they can be paid
more. At the same time, the slows would like to be confused with the quicks, so that they
can avoid being paid less. This is a general property of adverse selection models: the low-
er quality types generally attempt to associate themselves with the higher quality types,
who conversely attempt to set themselves apart from the lower quality types. Will the
quicks be able to signal their ability by obtaining a CPA?

In order for signaling to work, three conditions are necessary. First, if all other quicks are
signaling, and slows are not, an individual quick must be better off with signaling too. This
requires that pay net of the quick’s cost of the CPA is higher than pay if she decided to join
the slows by not getting a CPA:

Qq – Cq > Qs.

Second, if all other slows skip the CPA and quicks invest in the CPA, an individual slow
must be better off not getting the CPA as well. If a slow decided to infiltrate the quicks, he
would be paid Qq, but it would cost Cs. If he did not, he would earn Qs. Thus, for slows to
skip the credential, it must be true that:

Qq – Cs < Qs.

These two together imply that:

Cq < Qq – Qs < Cs.

Intuitively, the gain from signaling must be higher than the cost for high ability types, but
not so high that low ability types are also motivated to signal.

Third, for all of the quicks to be willing to signal, their profit from doing so must be higher
than what they would get if none of them signaled at all. If none signal, everyone is paid
average productivity, so this requires that:

Qq – Cq > Q ..

This condition is even stronger than the other condition for quicks described above, since
Q > Qs. Very large α implies that Q is very close to Qq, making it more likely that this last
condition cannot be met.

Intuitively, signaling to distinguish themselves from the crowd is more likely to be profitable
for quicks the rarer that they are. If there are many high quality types, it is relatively easy
for low quality types to hide themselves among the high quality group.

If these conditions are not met, neither has an incentive to obtain the credential, and
quicks do not distinguish themselves from slows. In this case there is no signaling. Such a
case is called a pooling equilibrium. This illustrates that signaling is not always possible.

If the conditions are met, quicks signal and slows do not. This is called a separating equili-
brium, since quicks are able to separate themselves out from slows by investing in the
credential.

34
 These points provide a formal illustration of arguments we provided about screening earli-
er. Signaling involves screening, but adds an important new ingredient: workers know their
type, and firms try to structure job offers so that those who are a better match for their firm
reveal this by their willingness to accept the offer, while those who are a poor fit reveal this
by refusing to accept the offer.

Which Type of Firm is More Likely to use Signaling?

Signaling is helpful when employers do not have enough information about job applicants
to assess their potential accurately enough. It is useful when differences in talent among
potential employees matter a lot to productivity. When differences in talent do not make
much difference to productivity, signaling will not be very useful. These ideas suggest
when we should expect to see employment practices consistent with signaling.

First, signaling should be more important in jobs where skills are most important. Such
jobs tend to be those that are at high levels of the hierarchy, in research and development,
and in knowledge work. They also correspond well to professional service firms, such as
consulting, accounting, law firms, and investment banks. In such professions, even small
differences in talent can lead to large differences in effectiveness on the job, so sorting for
talent is very important. For this reason, such firms tend to screen very carefully at recruit-
ing, and usually have promotion systems that correspond well to our probation story
above, at least in the first few years on the job.

Signaling is also more likely to be used where there is not much information already avail-
able about job applicants. Workers who are new to the labor market (say, having just
graduated from college or an MBA program) are more likely to see such policies. New
hires with many years of experience, and an extensive resume of past accomplishments,
should expect to see less use of signaling policies in the job offers that they receive. Nev-
ertheless, firms can use these techniques even for hiring experienced talent at very high
levels, when appropriate. For example, a new CEO is often hired for a fixed period con-
tract, with extensive pay for performance. To the extent that the CEO’s ability to implement
the strategy is hard to determine, and the CEO knows more about this than the board’s
hiring committee, such practices can improve CEO recruiting.

Summary

An important objective of personnel policies is to sort talent to appropriate employers and

positions in order to increase organizational effectiveness. Since there is incomplete and
often asymmetric information during recruiting, firms must screen job applicants. When a
high salary is offered, unqualified workers may apply for the job if there is a significant
chance that they can survive. This is the economic problem of adverse selection.

There are a number of ways that a firm can cut down the number of undesirable appli-
cants. One way is to look for credentials that are good predictors of on-the-job perfor-
mance. This works well when the credential is easy to obtain for qualified people but diffi-
cult to obtain for unqualified ones.

Beyond credentials, firms can invest in more or less extensive screening when hiring. This
can include formal testing, psychological profiling, lengthy and multi-round interviews, and
trying workers out on the job briefly. All of these can be helpful, though it is likely that they
will be far from perfect predictors of job performance.

35
The most accurate screen is to employ workers for some probationary period on the job.
Of course, it may also be the most expensive. The worker must be paid during this period,
and costs can be especially severe in jobs where there is a large downside risk in the
worker’s job (ability to destroy value). Many firms use some form of probation, either for-
mally or informally, in their hiring practices.

In our analysis workers were offered a contingent contract. In the case of probation, work-
ers receive sufficiently low pay during probation so that only those who believe that they
will be successful are willing to apply for the job. A well-crafted probation and post-
probation salary schedule can keep undesirable applicants from applying while attracting
desirable ones. This is easier to achieve when it is difficult for unqualified workers to sneak
past the probationary period and when qualified and unqualified workers have similar out-
side opportunities.

Screening and probation introduced the economic idea of signaling. Signaling is a method
that sometimes can solve adverse selection problems. If workers know their ability, quali-
fied workers would try to make this known to employers by signaling, and unqualified
workers would try to hide the fact that they are unqualified. Therefore, firms should design
recruiting policies to encourage qualified workers to apply, while making it difficult for un-
qualified workers to get through probation.

Another possible way to induce good self-selection among job applicants is to use strong
pay for performance. In fact, probation does this, since the post-probation promotion and
raise in pay is contingent on good performance. More generally, incentive pay of any kind
tends to improve recruiting, since better fitting employees are more likely to accept strong
incentive pay.

Firms can use contingent rewards after probation to attract higher-quality workers. How-
ever, this is not free; it comes at the cost of higher wages. These policies are most likely to
be useful in companies where small differences in talent can lead to large differences in
productivity of employees. This is most likely at high levels of the organization, and in firms
where intellectual work is emphasized. Professional service firms, especially leaders in
their respective industries, often employ policies of this type. These include careful recruit-
ing, extensive performance evaluation during the first few years of employment, and up-or-
out promotion systems with large rewards to those who are retained. They may also in-
clude strong individual pay for performance.

In the classic view of economics, goods are bought and sold on spot markets, with the
terms of the transaction consisting of quantity, quality, and price. Our analysis in this chap-
ter has opened up a different view: when firms use probation or contingent pay to screen
workers, they offer a multiperiod contract. This contract is contingent on the employee’s
performance: how they are treated depends on how they perform. Finally, it also involves
a promise by the firm – to reward those who perform well with higher pay later. These
complications arise because the quality of the good – ability of the worker – is not readily
available information. Thus, the economic relationship between the employer and the em-
ployee becomes complex. This idea is developed further in the next chapter, and even fur-
ther still in Chapter 15 at the end of the text.

36