You are on page 1of 28

It's Not What You Pay it's the Way that

You Pay it and that's What Gets Results:


Jockeys' Pay and Performance

Sue Fernie - David Metcalf

Abstract. Management scholars and economists have recently set out the
requirements of a system to elicit good performance when it is necessary to align
the interests of the principal and agent. We analyse pay and performance in an
occupation — jockeys — replete with moral hazard possibilities. We are able to
do this because, most unusually, a measure of pure individual performance exists
for an unbalanced panel of some 50 individuals for 8 years. Three hypotheses are
tested. First, in line with classic agency theory, we expect monitoring mechanisms
and incentive contracts to be used to align the interests of principals and agents.
Second, pay and performance should be positively associated, subject to the first
hypothesis being confirmed. Third, a limited number of jockeys were paid via an
alternative mechanism involving very large non-contingent retainer fees. This
serves as our counterfactual payment system. In line with agency theory we expect
worse performance under such a system than under an incentive contract. The
three hypotheses are confirmed: incentive contracts generate superior
performance to non-contingent payment systems. Our evidence suggests that
'it's not what you pay it's the way that you pay it... and that's what gets results'.

It is maddening that society confers its blessings on traditional academic


pursuits but views the study of horseracing as utter frivolity (Beyer, 1983).

Sue Fernie - David Metcalf, Centre for Economic Performance, London


School of Economics, Houghton Street, London WC2A 2AE, UK.
Steven Mclntosh and Stephen Woodland provided efficient research assistance
and constructive suggestions. We also benefited from comments at the Pay
Systems Study Group at the tJS Industrial Relations Research Association in
New Orleans, the LSE and Oxford labour seminars, and from discussions with
Stephen Dunn, Bob Gould, Paul Gregg, Brian Hall, Alan Manning, Morris
Perlman, Stephen Machin, Andrew Oswald, Jonathan Wadsworth, John Whitley
and Stephen Wood.

LABOUR 13 (2) 385-411 (1999) JEL J33, M12


© Fondazione Giacomo Brodolini and Blackwell Publishers Ltd 1999, 108 Cowley Rd.. Oxford 0X4 IJF, UK and
350 Main St., Maiden, MA 02148, USA.
386 Sue Fernie - David Metcalf

1. Introduction

Mechanisms to elicit good work performance, particularly in


organizations where it is difficult to measure effort, have long been
a focus of study. The potential for not working efficiently exists in
many jobs. The problem normally arises because the interests of
the principals (e.g. owners and managers) are not perfectly aligned
with those of their agents (e.g. the other employees). Industrial
relations scholars have examined how different payment systems
generate more or less effort or output (e.g. Brown, 1962; Flanders,
1970; Richardson, Wood, 1989; Roy, 1952), but it is only recently
that theorists (e.g. Jensen, Murphy, 1990; Lazear, 1995; Milgrom,
Roberts, 1992) have set out the principles underlying payment
systems to overcome the inherent moral hazard under principal-
agent relations in the labour market. This requires both monitor-
ing arrangements and contingent pay.
We analyse how members of one highly skilled, dangerous
occupation replete with moral hazard possibilities — jockeys —
are monitored and paid. Are the official arrangements consistent
with those put forward by agency theorists? We set out the
monitoring arrangements and incentive contracts and note that the
payment system also has an insurance component.
This leads to the following questions. If the monitoring and
payment system are congruent with the predictions of agency
theory, does this yield good performance? Do those with better
performance get higher pay? We test whether or not the official
jockeys' payment system elicits good performance. Our research is
confined to flat racing because corresponding performance
information is unavailable for national hunt Gump) jockeys.
It is important to emphasize that this paper provides a very
strong test of agency theory. First, we are able to examine whether
or not agents respond to different incentive systems because we can
compare performance under incentive contracts with performance
under a non-contingent system. Second, the strength of the pay-
performance association is estimated. Most previous literature on
this issue has examined pay and performance of chief executive
officers (CEOs). But generally such studies do not test agency
theory — they do not analyse how agents respond to different
incentives. Rather, they simply calculate the strength of links
between pay and performance, typically concluding until very
recently that, at best, the relationship is very weak.' A further
advantage of this study arises from our measure of individual
I Fondazione Giacomo Brodolini and Blackwell Publishers Ltd 1999.
Jockeys' Pay and Performance 387

performance which, most unusually, captures the contribution of


the agent to value added — it reflects pure individual performance
and is not contaminated by the influence of any other factors.
The remainder of the paper is organized as follows. Section 2
discusses agency theory, monitoring and contingent pay and sets
out the three hypotheses we wish to test. Background information
on the nature of the industry and the jockeys' labour market is
presented in Section 3. Section 4 describes the sample, sets out our
method of investigation and discusses the pay and performance
variables. Our results are presented and discussed in Section 5 and
conclusions are given in Section 6. Full details of the calculation of
performance are given in Appendix A.

2. Theory and hypotheses

Moral hazard is behaviour under a contract that is inefficient


and arises from the differing interests of the contracting parties. It
persists only because one party to the contract cannot tell for sure
whether the other is honouring the contract terms. Moral hazard
problems arise frequently in principal-agent relationships, where
the agent is called upon to act on behalf of the principal, because
the agent's interests commonly differ from the principal's and the
principal cannot easily evaluate how well the agent has worked or
whether the agent has been honest.
Classic agency theory points to two main ways to control moral
hazard problems (Milgrom, Roberts, 1992). First, monitoring can
lessen the information problem that is a fundamental component
of moral hazard. Monitoring requires developing sources of
information about the agent's truthfulness and performance. But
sometimes monitoring actual behaviour or the truthfulness of
reports may be too costly. Therefore, second, incentive contracts
can be used such that pay is for outcome performance — results
— when inputs cannot be measured. However, outcomes are not
fully under the control of the agent, and risk aversion on the part
of employees therefore implies that not all pay will be outcome-
related, i.e. the workers (agents) will not wish to bear the full risk
concerning the variability of earnings of the organization. An
efficient incentive contract therefore involves balancing the costs of
risk bearing against the benefits of improved incentives: 'insulating
risk-averse employees' pay from variations in measured job
performance minimizes the costs of risk bearing, but it also
© Fondazione'Giacomo Brodotini and Blackwell Publishers Ltd 1999.
388 Sue Fernie - David Metcalf

eliminates monetary performance incentives. Shifting risk to the


employees strengthens their incentives because their pay now
depends on actual performance, but the costs of risk bearing rise as
weir (Milgrom, Roberts, 1992, p. 188). A well-designed incentive
contract achieves 'goal congruence' removing any conflict of
interest by effectively altering individual objectives, such that those
of the agent are now ahgned with those of the principal.
We show below that the occupation we are studying — jockeys
— abounds with moral hazard possibilities because, absent
monitoring and incentive contracts, the interests of principals
(owners and trainers) are imperfectly aligned with those of their
agents (jockeys). Therefore, as predicted by classic agency theory,
our first hypothesis is that extensive monitoring of the agent's
behaviour occurs, but as such monitoring is inevitably costly and
imperfect, an incentive contract will also be used. In line with
agency theory we hypothesize that this contract will incorporate
both performance related and insurance elements.
Our second hypothesis flows from the first. Subject to agency
theory being confirmed under our first hypothesis, we expect pay
and performance to go hand-in-hand. Without incentive contracts
and monitoring we would expect no association between pay and
performance. It should be emphasized that any such link is not
tautological (unlike associations found in some studies of CEOs
which define total remuneration as base pay plus the value of share
options and define company performance by the share price). The
measure of performance we use is calculated completely indepen-
dent of pay. It is an objective individual performance indicator
which reflects the employee's contribution to the value of the firm.
This is unusual evidence and our second hypothesis therefore
provides superior evidence regarding the agency model than is
normally the case.
The third hypothesis concerns our counterfactual. In cases where
the payment system differs from that predicted by classic agency
theory — for example where payment is not contingent on
outcomes — we expect performance to be inferior to that generated
by properly designed incentive contracts. In order to test this
hypothesis a counterfactual payment system is required. British
jockeys have a long history of being paid according to prize money
won and the number of rides (Munting, 1987), and similar systems
operate in other countries with horseracing. However, we are
fortunate because our sample does embrace an alternative payment
system and therefore permits a contrast between performance
© Fondazione Giacomo Brodolini and Blackwell Publishers Ltd 1999.
Jockeys' Pay and Performance 389

outcomes under agency theory and those under an alternative


system where the agent bears none of the risks yet receives
handsome rewards. This latter system has elements of the shirking
model: a fixed wage which the individual continues to receive as long
as performance is up to scratch. Thus we are able to examine
whether agency theory dominates shirking theory or vice versa.

3. The horseracing industry

The horseracing industry extends from breeding to betting and


400,000 people are economically dependent on the industry (Home
Affairs Committee, 1991). The organization and administration of
racing is conducted by the British Horseracing Board (BHB) and
the Jockey Club. They co-ordinate racing's activities with the
financial support of the Levy Board, owners and racecourses. The
BHB provides the necessary expertise to protect the integrity of the
racing product.
Around 11,000 thoroughbreds are born each year in Britain and
Ireland as a result of planned matings between some 1,000 stallions
and 17,000 mares. The system is one of arranged pairings and
dowries. The breeder — owner of the mare — chooses a mate to
suit the mare and pays a fee, ranging from £120,000 for supersire
Sadlers' Wells to under £1,000 for a lowly jump race stallion
(Montgomery, 1995; Pacemaker and Thoroughbred Breeder,
1995).
Nearly 9,000 owners had horses in training in 1995. They are
racing's major financiers and biggest risk takers, supporting
trainers and their employees, as well as the breeding industry.
Racehorse owners' overall deficit in the 1990s (see, for example,
Jockey Club and Horse Racing Advisory Council, 1991) has been
over £150 m each year. Owners' costs are approximately £200 m
yet the prize money received is only some £50 m. Owners' losses
presumably reflect the enjoyment and other non-pecuniary benefits
of owning racehorses. The owner normally hires an agent to train
his or her horse. There were nearly 600 licensed trainers in 1995,
employing around 6,000 stable staff.
In 1995, 639 flat race fixtures were distributed among 35
racecourses. The race programme provides a minimum of six races
per fixture, constructed with the objective of providing a regular
supply of attractive, competitive races for betting purposes and
which also meet the needs of the horse population. In 1995, 4,154
© Fondazione Giaeomo Brodolini and Blackwell Publishers Ltd 1999.
390 Sue Fernie - David Metcalf

fiat races were run. The total number of runners was 41,692, an
average of 10 per race.
Consumers of racing's product range from those pursuing the
sporting and entertainment aspects to those interested in the
speculative element of horseracing. Total public attendances at
racecourses in 1995 were 4.7 milhon. Off track there were 9,670
betting shops employing around 50,000 people.
In 1995 there were 109 licensed jockeys and a further 201
licensed apprentices. A potential jockey must be employed by a
licensed trainer as an apprentice. Their tasks are similar to those of
a stable lad, including mucking-out and riding work, but they also
get to ride in races. Few apprentices become full jockeys. In 1995
there were 109 full fiat jockeys. Assuming (conservatively) an
average working life of, say, 10 years, 11 would leave each year.
Thus, assuming a 4 year apprenticeship, some 50 apprentices are
essentially competing for 11 slots — nearly five per slot. The
unsuccessful aspirants will either leave the industry or downgrade
their expectations and rest content with work riding and other
stable tasks. Normally when the apprentice is promoted to be a full
jockey he becomes self-employed ('freelance'), although some are
attached to ('retained by') a particular stable or owner.
The supply of jockeys is deliberately limited. The occupational
licence is issued by the BHB only to those who have fulfilled the
requisite apprenticeship. The full licence costs £120 while the
apprentice licence costs £60. The rules concerning licences run to
five and a half closely typed pages in the Rules of Racing and
Instructions (British Horseracing Board and Jockey Club, 1993).
Riders from overseas are a potential alternative source of
supply. Not surprisingly, 'the Jockeys Association is concerned
that overseas riders could be taking valuable opportunities away
from British riders' {The Independent, 3 March 1993). A rider from
the EU (of whom there are very few) can be granted a temporary
30 day permit by the Jockey Club. A non-EU rider from overseas
would, in addition, require a work permit issued by the Employ-
ment Department. In essence, star riders from overseas are
permitted to ride here but workaday performers are not. This is
similar to rules governing employment contracts in many other
fields. Such international stars clearly add spice to the scene: in the
last decade the champion fiat jockey has been from abroad four
times.
Given the labour supply, it is in jockeys' interests if the demand
for their services is inelastic. This depends on the elasticity of
© Fondazione Giacomo Brodolini and Biackwell Publishers Ltd 1999.
Jockeys' Pay and Performance 391

demand for the product, the ease of substituting other factors for
jockeys and labour costs in total costs. Each of these Marshallian
rules are favourable to jockeys.
Despite rising relative costs of ownership and attendance the
number of runners has been roughly constant in recent years at
some 40,000, and total attendances have remained within the 4.5-5
million bracket. This hints that the demand for the product is
inelastic.
It is impossible to substitute capital for jockeys' services (though
punters sometimes think it would be preferable!) and supply is
limited, as we saw above, by the licensing system. The substitution
possibilities occur only within the given pool such that an owner or
trainer can switch jockeys at will.
Jockeys are fortunate that their labour costs are a very small
fraction of total costs — 'the importance of being unimportant'. In
1995 jockeys' riding fees totalled between 2 and 3 percent of the
costs borne by owners for keeping horses in training. Even though,
in addition, jockeys receive a share of the prize money and are
sometimes on retainers, the cost of the jockey is not a major
consideration in an owner's decision to enter, remain in or leave
the industry.
The Jockeys' Association (JA), while not a union as defined by
the Certification Officer, fulfils normal union functions of
providing a collective voice in matters like safety and discipline,
individual representation and undertaking pay and other negotia-
tions with counterpart institutions such as the Racehorse Owners'
Association and BHB. The Association is financed by a levy of
£1.45 per ride from the jockey's riding fee. The JA has a difficult
task reconciling both the interests of racing and jockeys and of
accommodating different jockeys' concerns. Among jockeys, the
interests of the workaday all-weather rider are dissimilar from
those regularly riding at Ascot and Newmarket.
In the absence of monitoring and incentive contracts, there is
ample scope for moral hazard in racing. The principal (trainer)
cannot easily gauge the effort, quality and commitment of the
agent (jockey). It is very tricky for the principal to verify the
information provided by the agent. The jockey might report, for
example, that the horse does not like the going (e.g. ground too
soft) or a left-handed track or does not get the trip. He could do
this simply to mask a poor ride or, alternatively, because he has
deliberately not tried, perhaps because he has accepted a 'present'
from a bookmaker with substantial liabilities on the horse, or
© Fondazione Giacomo Brodolini and Blackwell Publishers Ltd 1999.
392 Sue Fernie - David Metcalf

perhaps in order to avail himself of lengthened odds on the horse


next time it runs. The Chairman of the Jockey Club Disciplinary
Committee summarized the problem thus {Racing Post, 2 March
1995): 'In any sport in which one of the greatest motivators is
accrual of money through betting and prize money, there is always
going to be the temptation, as it were, to fix the odds your way. So
there are bound to be, always have been, always will be, people
who will try and get around the system.' The use of incentive pay
systems to overcome moral hazard in the face of imperfect
monitoring is set out in Section 5.

4. Data and method

We investigate links between pay and performance for a sample


(unbalanced panel) of those jockeys for whom the performance
measure is calculated each year for the period 1988-95 (« = 413).
The frequency with which each observation appears in the sample
is:

Years in sample 1 2 3 4 5 6 7 8
Frequency («) 18 14 6 8 5 4 11 23

We also present some evidence from the balanced panel of 23


jockeys who were in the sample in all 8 years {n — 184). In addition
we analyse performance over an extended period (1983-95) for
those jockeys who received large non-contingent retainer fees for
some part of the period {n — 72).
The three hypotheses set out in Section 2 are tested as follows.
First, the payment system is described to see whether it is
consistent with that suggested by agency theory. This requires
elements of both insurance and an incentive contract. Trade-offs
between monitoring and incentives are also examined.
The second hypothesis is that, provided the payment system is
consistent with agency theory, pay and performance go hand-in-
hand. This is analysed by a random effects longitudinal regression
of pay on performance for our sample. The random effects model
is appropriate, rather than ordinary least squares, because it
accounts for unobserved heterogeneity among jockeys which might
otherwise bias the results. We also estimated a fixed effects model,
© Fondazione Giacomo Brodolini and Blackwell Publishers Ltd 1999.
Jockeys' Pay and Performance 393

with a dummy variable for each jockey. The Hausman test (see
Green, 1991) indicates that the random effects model is preferred
to the fixed effects model and therefore the latter is reported in
Appendix C rather than in the text, together with results using
ordinary least squares.
Third, for a limited number of individuals, one element in the
payment system changed dramatically during the period. A few
mega-rich owners retained a particular jockey to ride their horses.
The sums involved were very large (up to £1 m a year) and — in
stark contrast to other elements in the payment system — no
performance criteria were specified in these arrangements. We
analyse the performance of each such jockey when retained and
when not retained. Our third hypothesis is that performance when
in receipt of such non-contingent largesse is inferior to perfor-
mance under the normal incentive contract. We now turn to
provide more information on the performance and pay variables.
In most occupations and workplaces, performance is measured
either on the basis of objective outcomes (e.g. profits) or by
subjective assessments of supervisors (e.g. merit pay). Outcome
measures commonly used for jockeys are total wins, total prize
money and the strike rate (wins/rides). These measures are self-
evident but do not refiect performance perfectly because none
controls for the crucial input — horse quality.
An alternative is the return to level stakes (say £1) on all
mounts, i.e. the aggregate returns over a season from placing a, £1
win bet on each horse the jockey rides. Unfortunately, this measure
also fails to accurately measure jockeys' performance. There are a
number of reasons for this. First, shorter prices are less weighted
against the backer than long ones, i.e. the odds of the favourite
(shorter priced) in a race more accurately refiect its chance of
winning than do the odds of an outsider (longer priced). Given
that a fashionable jockey will ride a higher proportion of short
priced horses than an unfashionable jockey, t;heir respective profit/
loss figures will be biased by the above effect. Second, if the jockey
rides for a serious gambling stable the occasional coup will boost
his returns to level stakes. Third, if many of the jockeys' rides come
from a trainer having a bad season because a virus is causing
sickness among the horses in the stable, his returns will tend to be
depressed because prices tend to lag the stable's form. Finally, if a
rider is popular (or unpopular) with the general public the returns
are affected — the public 'overbet' a popular jockey thereby
depressing the price on his mounts.

1 Fondazione Giacomo Brodolini and Blackwell Publishers Ltd 1999.


394 Sue Fernie - David Metcalf

'The truth is that jockeys' relative abilities may be quantified


only by exhaustive examination of each horse's performance under
different partners through a season' (Millan, 1995). This is
precisely our preferred measure. It fulfils the requirement of Baker
et al. (1994) that 'an ideal performance measure would reflect an
employee's contribution to firm value'. Such pure performance
measures hardly ever exist because, as Baker et al. write, 'an
employee's contribution to firm value usually is not objectively
measurable'. In a similar vein, Stiglitz (1991) notes that 'the
problem is that only some attributes of output can be measured
accurately or at all'. Our indicator is an exception to this general
rule — it captures pure performance.
Information concerning the Racing Research performance
indicator is given in Whitley (1992) who writes that
The only logical way to measure a rider's effectiveness is to examine
how horses have performed for him compared with their performances
for other jockeys. This is what we do. The analysis is confined to riders
for whom there is a large amount of data and this data is analyzed in a
mathematically correct way by computer programs. The resulting
ratings, given in pounds to the nearest tenth, are therefore an accurate
reflection of each jockey's effectiveness over the season. In particular,
if a rider is the best jockey then horses in general should run better for
him than for other jockeys. And if they do he would appear at the top
of our ratings.
Whitley's method used to calculate his (and our) performance
indicator is set out in Appendix A. It is a cardinal measure
(calibrated for convenience to have a mean of 10 lb). It is
normally accepted that over 5 furlongs (approximately 1 km),
other things being equal, a horse carrying 3 lb more will be a
length worse off. This true performance indicator can differ
markedly from public perception. For example, the 1993 official
champion jockey Pat Eddery — as defined by number of wins —
was rated only 19th, over 2 lb worse off than the top-ranked
jockey, on our measure.
The official payment system (British Horseracing Board and
Jockey Club, 1993) permits a calculation of-the pay for each jockey
for each year:
Pay = a (prize money) + (number of rides x fixed fee).
The fraction of prize money {a) and the riding fee are the same
for all jockeys. Prize money distribution is set out in Table 1. The
© Fondazione Giacomo Brodolini and Blackwell Publishers Ltd 1999.
Jockeys' Pay and Performance 395

Table 1. Distribution of prize money (%

All pattern All races Other races Other races Average


races with with six with four with only
four prizes prizes prizes three prizes

Winner
Owner 44.44 44.89 50.11 53.93
Trainer 5.78 5.59 6.24 6.74
Jockey (b) 4.06 3.92 4.38 4.76
Stable 3.00 2.90 3.25 3.50

Total (a) 59.28 57.30 63.98 68.93


Jockey/total win 6.85 6.84 6.85 6.91 6.86

Place
% of total* 39.32 41.30 34.62 29.67
(all parties)
% of total to 1.77 1.91 1.45 1.24
jockey (c)
% of place to 4.50 4.62 4.19 4.18 4.37
jockey

% of total to 5.83 5.83 5.83 6.00 . 5.87

Notes: (i) * = 100 - (a + 1.40). The 1.40% refers to 0.60% for apprentice training, 0.20%
for the jockey valet fund and 0.60% for the Jockeys' Association pension fund,
(ii) Stable in the first column refers to share of prize money going to the stable
employees (excluding the jockey).
(iii) Pattern races (column 2) refers to the ultimate 107 top class races, which have
large prize money. For example. Group 1 races for 3 years olds and up have
minimum prize money of £95 000 and those for 2 year olds minimum prize money of
£60000.
Source: Calculated from British Horseracing Board and Jockey Club (1993), Rules 191-199.

jockey receives 6.86 percent of win prize money and 4.37 percent of
place prize money.^ The number of rides depends on the jockey's
reputation and gets multiplied by a fixed riding fee (presently,
£61.50). Each year the coefficient of variation (standard deviation/
mean) of prize money earnings (typically approximately 1.0) is
around three times greater than the coefficient of variation of
riding fees (typically 0.3).
The remaining element in the payment system is large retainer
fees received in some years by a select sample of jockeys retained to
ride for owners from the royal families of Dubai and Saudi Arabia.
Details are given in Table 3. An annual retainer fee of £0.5 m is
© Fondazione Giacomo Brodolini and Blackwell Publishers Ltd 1999.
396 Sue Fernie - David Metcalf

some 10 times larger than the annual amount the average top 50
jockey would receive from prize money earnings plus riding fees.

5. Results and discussion

5.7 Hypothesis 1
Our first task is to see whether the monitoring and incentive
payment schemes are consistent with agency theory. We show
that racing has developed an array of measures to combat the
incentive to moral hazard behaviour. First, there is the detailed
monitoring of the agent's performance. Second, the monitoring is
imperfect so the payment system involves an explicit incentive
contract which is a textbook example of the use of contingent pay
systems to signal the correct incentives. Racing provides a fine
example of Herbert Simon's (1991) statement that moral hazard
can be overome 'via monitoring combined with contracts that
appeal to self-seeking nature'. Agency theory also notes that
agents tend to be risk averse and that payment systems must
allow for that through an insurance element such that pay is not
wholly performance related. This, too, is found in the case of
jockeys.
There are many sources of monitoring jockeys' performance in
the 40,000 total annual rides but few will be completely accurate.
The trainer or owner may believe that the jockey underperformed
and occasionally may publicly criticize the jockey. But more subtly,
and more normal, the trainer or owner may simply quietly refuse,
perhaps temporarily, to employ a particular jockey.
Formal monitoring of jockeys' performance is undertaken by a
third party — the regulatory body, the Jockey Club. Video
recordings for the surveillance of racing by the Jockey Club are
central to such monitoring. Further, there are both general rules
and specific rules on the day that jockeys must abide by. Crucially,
jockeys are not allowed to bet. Rule 62 ii (British Horseracing
Board and Jockey Club, 1993) prohibits jockeys from owning a
horse, betting (or instructing someone else to bet) on a horserace,
receiving the proceeds of a bet on horse racing, and receiving
presents in connection with a race (other than from the owner of
the horse he rides in a race). Rule 204 iv prohibits a jockey from
passing information concerning a horse's prospects (perhaps to a
bookmaker) in return for 'monetary consideration'. Such rules
© Fondazione Giacomo Brodolini and Blackwell Publishers Ltd 1999.
Jockeys' Pay and Performance 397

encourage jockeys to ride each horse on its merits: if obeyed, they


cannot gain today or in the future from not trying.
Stewards at the racecourse deal with discipline on the day. Their
concerns revolve around the jockey either not trying, or trying too
hard, and misusing the whip. For around 1 ride in 100 the local
stewards examine whether or not the jockey was fully trying. Not
trying is a very serious offence (Rule 151), which will also involve
the trainer and owner. Approximately 1 ride in 1,000 results in
suspension, recording of the explanation or referral to Portman
Square (Jockey Club headquarters).
Punters at the track also provide a useful form of third party
monitoring. Further, nearly all races are now televised both in
betting shops and via home TV subscription channels, thereby
increasing the pool of punters who can see for themselves the
behaviour of horse and jockey.
Monitoring by trainers, owners, the Jockey Club and punters
limits moral hazard behaviour in the principal-agent relationship.
If the jockey's performance does not satisfy the trainer and owner
he will lose mounts from those principals in the future. If he comes
to the notice of the Jockey Club by being involved in an attempt to
deceive the handicapper or set up a betting coup the penalties are
severe, ultimately being warned off (banned from any involvement
in racing).
However, the monitoring can never be perfect. Hence the
payments system for jockeys is designed to further reinforce proper
behaviour. The official payment system for jockeys is a classic
example of using contingent pay to overcome potential moral
hazard problems:
Pay = a (prize money) + (number of rides x fee).
The fraction of prize money {a) and the riding fee are constant
across jockeys. The pay system therefore involves two elements.
First, an explicit incentive contract relating to prize money won.
Second, a fixed fee per ride, providing insurance for those who win
little prize money.
The incentive contract is used to promote effort and diligence. It
motivates good behaviour by rewarding superior outcomes. By
relating pay to prize money won the jockeys gain by getting the
principals what they too desire. Full details are set out in Table 1.
The distribution of prize money is virtually identical for all races
so the elasticity of jockeys' earnings with respect to prize money is
approximately unity: a doubling of prize money yields a doubling
) Fondazione Giacomo Brodolini and Blackwell Publishers Ltd 1999.
398 Sue Fernie - David Metcalf

of pay. This achieves another important ingredient of the


theoretical ideal: the larger the absolute return to the principal
(owner),, the larger the incentive to the agent (jockey) to try his
best. The jockey gains more for winning the Derby than he does
for winning a class F race at Folkestone. It should be noted,
however, that the incentive carrot is backed by a regulatory stick.
Under Jockey Club Rule 151a horse must always be ridden for its
best possible place so the jockey must also perform to his best at
Folkestone.
Our first hypothesis is confirmed. In the face of imperfect
monitoring arrangements a very sensitive incentive contract is used
to align the interests of principal and agent. However, to allow for
risk aversion by agents the incentive contract is modified by a fixed
element in the pay system.
It is interesting to compare the share of revenue (prize money)
accruing to a jockey for winning a race (set out in Table 1) with
corresponding CEO and managerial incentive contracts. The
winning jockey receives some 7 percent of the total revenue and
the elasticity of pay with respect to revenue is unity. By contrast,
the most thorough study of executive compensation in the UK
(Main et al., 1996) puts the elasticity of CEO pay with respect to
sales at 0.13 for base pay and 0.33 for base pay plus the imputed
value of share options. Similar orders of magnitude are also
reported, for example, by Conyon et al. (1995) for the UK and
Jensen and Murphy (1990) for the USA. However, recent US
evidence suggests that CEO incentive contracts are much more
high powered than previous studies suggest. Hall and Liebman
(1997) state that when CEO holdings of stock and stock options
are included in compensation the median elasticity of CEO
compensation with respect to firm value is 3.9 for 1994.

5.2 Hypothesis 2
We now turn to examine our second hypothesis, namely whether
or not this payment system causes pay and true individual
performance to go hand-in-hand. The evidence is set out in
Table 2 which is the heart of the paper. Racing Research
performance is regressed separately against total pay and its two
constituent components: riding fees and prize money earnings.
Information is also presented separately for the unbalanced panel
and the balanced panel of 23 jockeys who are in the sample in all 8
years. The results using the balanced or unbalanced panel samples
© Fondazione Giacomo Brodolini and Blackwell Publishers Ltd 1999.
Jockeys' Pay and Performance 399

Table 2. Current pay (£) regressed on current Racing Research


performance, 1988-95

Sample, pay definition Regression coefficient Constant R^

Unbalanced panel, levels 5,733 -18,074 0.20


(5.3) (-1.6)
Unbalanced panel, logs 0.92 8.27 0.28
(6.6) (26.0)
Balanced panel, levels 5,763 -6,201 0.19
(2.6) (-0.05)
Balanced panel, logs 1.01 8.36 0.29
(3.7) (13.2)
Riding fees only
Unbalanced panel, levels 1,684 1,502 0.28
(6.0) (0.5)
Unbalanced panel, logs 0.64 8.33 0.31
(6.8) (39.0)
Balanced panel, levels 1,375 7,922 0.34
(2.8) (1.5)
Balanced panel, logs 0.58 8.61 0.38
(3.6) (22.9)
Prize money earnings only
Unbalanced panel, levels 4,117 -20,052 0.18
(4.6) (-2.1)
Unbalanced panel, logs 1.32 6.43 0.30
(6.4) 13.5)
Balanced panel, levels 4,502 -10,290 0.17
(2.3) (-0.5)
Balanced panel, logs 1.55 6.32 0.33

Notes: (i) All regressions use a random effects model and have Racing Research on the
right-hand side and pay on the left-hand side,
(ii) Observations: unbalanced panel 413, balanced 184.
(iii) f-statistics are in parentheses.
(iv) Time dummies included (to control for increase in all nominal earnings),
(v) Definition of total pay is riding fees plus prize money,
(vi) Performance measure is Racing Research indicator; see Appendix A.

are similar. The evidence suggests that a jockey who currently


performs 1 lb (or 10 percent as calibrated in the Racing Research
data) better than another receives some £5,700 extra pay. The
elasticity of total pay with respect to performance lies between 0.9
and 1.0 and that for the incentive element alone is around 1.4.
Thus the sensitivity of pay to performance for jockeys is larger
than that found in the most recent UK study linking CEO pay and

© Fondazione Giacomo Brodolini and Blackwell Publishers Ltd 1999.


400 Sue Fernie - David Metcalf

performance: Main et al. (1996) define company performance by


share performance and include the value of share options in total
remuneration, yielding a corresponding elasticity of 0.72. It should
be noted, however, that their estimate is over twice as large as any
other for CEOs. It is unlikely that the lower sensitivity of pay with
respect to performance for CEOs in the UK occurs because it is
easier to monitor their efforts.
When total pay is divided into its constituent parts we see that
prize money earnings are around twice or three times as sensitive to
a given change in performance as riding fees. The incentive
contract whereby jockeys share revenue really bites: it generously
rewards those with superior performance and harshly penalizes
inferior performance.
It is also possible to examine whether or not reputation lags
performance. When the regressions reported in Table 2 were
estimated lagging performance initially by 1 year and then by 2
years, the sensitivity of pay with respect to performance rose over
time (these estimates are available on request). For the unbalanced
panel (row 1, Table 2), for example, a 1 lb (10 per cent) increase in
performance this year yields a £5,700 increase in total pay in the
current year, £6,000 extra in a year's time and £7,000 2 years on.
Thus the long-run elasticity, measuring the increase in the present
value of future earnings associated with better current perfor-
mance, is larger than the 0.9 to 1.0 calculated for current earnings
alone. So reputation does lag performance: it takes time for good
performance to get recognized, yielding more and better quality
future rides.^

5.3 Hypothesis 3

The vast bulk of jockeys are paid according to the official system
analysed above. We have shown that such a system is consistent
with classic agency theory and results in a strong positive link
between pay and performance. We are fortunate because we can
now compare this official payment system with an alternative
system based on huge non-contingent sums. During the last decade
large retainer fees have been paid to selected jockeys by owners
from the royal families of Dubai and Saudi Arabia. These wealthy
owners have horses with more than one trainer so the jockey rides
horses from more than one stable but is also free to ride for other
owners when the retaining owner does not have a runner.
I Fondazione Giacomo Brodolini and Blackwell Publishers Ltd 1999.
Jockeys' Pay and Performance 401

At least three important effort-performance-pay questions are


raised by these links. First, are the retained jockeys the best
available? Second, how does jockeys' performance compare with
and without these very substantial non-contingent payments?
Finally, does the retainer give good value to the owner? Does the
extra prize money flowing from the efforts of the retained jockey at
least cover the cost of the retainer? These are considered in turn. It
is not possible to conduct statistical tests on this evidence because
of the limited number of observations. Nevertheless, the data
provide a useful description of the consequences of non-contingent
payments.
Information on the cardinal and ordinal performance of the
seven main retained jockeys by the owners from the royal families
of Dubai and Saudi Arabia is given in Table 3. It covers 1983-95
inclusive but not all jockeys rode for all 13 years. The evidence in
Table 3 shows that only two jockeys — Dettori and Roberts —
were in the top 10 in both years prior to achieving the retainer. In
contrast Carson, Munro and Swinburn can, on this evidence,
count themselves lucky to have been chosen for such largesse. We
also present evidence on absolute performance: an average
performance is calibrated at 10 lb. All deviations from this
benchmark are positive: it is reassuring that the retained jockeys
had above average performance both when retained and not
retained. Again, however, the absolute performance indicators
hardly suggest Carson and Swinburn to be superstars.
Next, consider the second issue: the difference between the
performance figures for the retained years and the non-retained
years. In four cases — Carson, Cauthen, Eddery and Munro —
performance was unambiguously worse when retained than when
not retained. Only in two cases — Roberts and Dettori — was
performance when retained superior. These results are rather
striking. The retainers were worth up to £1 m per year yet, in the
majority of cases, instead of giving a boost to overall performance,
retainers had the opposite effect. It was only in the case of Roberts
and Dettori that their (already excellent) performance improved
still further. It is ironic that Roberts lost his retainer after only 1
year. Roberts' retainer (1993) was with Sheikh Mohammed, the
largest British owner. He had taken over from Cauthen (1991,
1992) when the latter had refused to take a cut in his fee, said to be
£1 m per year. Being retained jockey to Sheikh Mohammed was
described by those in racing as 'very difficult...just too many
horses and too many trainers' (Roberts, Tanner, 1994, p. 176),
© Fondazione Giacomo Brodolini and Blackwell Publishers Ltd 1999.
402 Sue Fernie - David Metcalf

•s
00 1/^ fN t-; Tf l o oq 3 2 3
01 fN —^ '-.' O — —' .c
•o .S

O "
.3 !a
ca o «
• PH f^ r^ ON m —' fN
fN — fN fN —' fN
06 •53 S
on "
Reta ined

o O "o VO O oo
— fN fN — fN o
Z O

Q g

•s Z 2
ill i n 1/^ O Tf \ o <N
O -^ fN —• —' fN —'
o.
o. 3
O
o
X

i =1
it
•s
.S 2

•s

I c V I — td ca

o
es
B6

.S « 2

« 2 S.2i
'i - .« 1-; u y
= 2
oq '
.11
<L> ••C « t. e
a ca

; 3 T3 3 •
|5>

o
i-j
iiiiiii
a a ij -a ^ o >

© Fondazione Giacomo Brodolini and Blackwell Publishers Ltd 1999.


Jockeys' Pay and Performance 403

inevitably resulting in the jockey oeeasionally ehoosing the wrong


horse for a particular raee. But evidenee here suggests that Roberts
did an outstanding job and was very unfortunate to lose his
retainer.
All in all our data strongly suggest that these large non-
eontingent retainer payments reintrodueed moral hazard into a
payment system whieh had previously proved to be very successful
at overcoming such behaviour.
The remaining issue concerns the value the owner gets from the
retainer system. Leaving aside non-monetary elements — a
brilliant ride by the retained joekey yielding a visit to the winner's
enclosure at the Derby or Royal Ascot — the question is: does the
extra prize money plus extra stud value of the horses generated by
the retained jockey equal or exceed the retainer? It is impossible to
do full justiee to this question because prize money won depends
more on the quality of the horse than the jockey. Nevertheless, the
answer seems to be negative.
Table 3 provides, for information, prize money won (1988-95
in £1,995) by our seven jockeys during the years they were retained
and not retained. Four joekeys had higher prize money when
retained, though only Dettori boosted his prize money by over
£1 m, but three — Carson, Cauthen and Swinburn — amazingly
amassed lower prize money when retained than when not.
But it is better to examine the matter from the owners' stanee.
Consider 1993, the last year before this owners' retainer system
began to crumble. The two top owners were Sheikh Mohammed
and Khaled Abdullah:

Sheikh Mohammed Khaied Abdullah

Total prize money won (£m) 2.6 1.6


Approximate amount to owner (£m) 1.3 0.8
Estimate of retainer (£m) 0.5 0.6

In the case of Khaled Abdullah, almost all of the prize money


accruing to him (in his best ever year) went in paying Eddery's
retainer. And even winning-most owner Sheikh Mohammed paid
over one third of his winnings to his retained joekey. It is surely
inconceivable that the extra prize money gained by these jockeys'
skills offset the cost of the retainer: so it is not surprising that this
© Fondazione Giacomo Brodolini and Blackwell Publishers Ltd 1999.
404 Sue Fernie - David Metcalf

system has decayed. Khaled Abdullah, Fahd Salman, Maktoum Al


Maktoum and Hamdan Al Maktoum have all axed their own
retained jockeys. And Sheikh Mohammed now has a looser system
via his trainers.
In one sense these large retainers are analogous to the wage
system under the shirking model where the worker is guaranteed a
fixed wage provided his performance is above a specified level. The
joekeys wish to keep their retained status and should try hard.
However, the evidence refutes the shirking model. The perfor-
mance criteria were not specified. The jockeys mainly performed
worse when receiving the large fixed payments — sometimes for
many years — than they did under the offieial incentive contracts.
Thus the shirking model is dominated by the more performance-
related contract and the former has withered away.^

6. Conclusions

Pure performance indicators — which measure 'the employee's


contribution to firm value' — are most unusual. For example,
much of the debate on CEO's pay is clouded beeause it is difficult
to measure their performance: outcome measures such as profits or
share values may have little to do with bosses' performance. We
chose to study jockeys because a pure performanee measure does
exist for this occupation.
We demonstrate that: (i) the official payment system for joekeys
is eoineident with the predietions of agency theory; (ii) under this
payment system earnings and performanee go hand-in-hand; and
(iii) the alternative payment systems — huge non-contingent
retainers — result in deteriorating performanee.
Economic and management theorists state that to elicit effort
and good performanee in principal-agent relations a payment
system must be based on an incentive contract. This is provided by
the prize money percentages. This system, eoupled with comple-
mentary monitoring arrangements designed to overcome moral
hazard, generates good performance. Indeed, the elastieity of total
current pay with respect to performance (approximately unity) is
mueh larger than that found in conneetion with CEOs' and
directors' pay in the UK. For example, Gregg et al. (1993) put the
elasticity at 0.5 (and, as they eannot measure pure performance,
this is probably an overstatement). Likewise, Main (1995), after
surveying all the British evidence, states that there is a 'surprisingly
© Fondazione Giacomo Brodolim and Blackweil Publishers Ltd 1999.
Jockeys' Pay and Performance 405

low pay-performance relationship at the top of large companies'.


And Main et al. (1996) also report a lower sensitivity of pay to
company performance for CEOs than is found in this study.
Further, if the present value of earnings is used to calculate the
elasticity the figure would be substantially higher. This confirms
the importance of payment systems — 'the way that you pay it' —
in eliciting good performance.
Alternative payment systems would be unlikely to deliver such
good performance. Consider some examples. First, when moral
hazard is present the principal can do the job him or herself. Gains
from specialization preclude most principals (owners) piloting
their own horses! Second, the jockey could keep ail the prize money
and pay the trainer/owner for the ride. Information asymmetries
concerning the state of the capital (horse) rule out such an option.
A weaker, third alternative would be for jockeys' pay to be entirely
based on prize money. This is ruled out by a combination of risk
aversion on the part of jockeys and the operation of the capital
market which would be unlikely to permit borrowing against
future earnings during a prolonged lean spell. Fourth, by contrast,
pay could simply be based on the number of rides (multiplied by a
given riding fee). Even if monitoring could be done properly, such
a mechanism would attract less productive performers. Further, we
have seen that reputation lags performance which also makes such
a payment system unsatisfactory.
The one strand of the payment system which appears not to have
delivered on performance is the huge retainers paid by the owners
from the royal families of Dubai and Saudi Arabia — 'it's not
what you pay'. The evidence suggests that, in the main, the initial
award of the retainer did not go to the best performing agents and
that jockeys receiving such largesse subsequently performed less
well than when they were not so retained. This finding suggests
that rupturing the link between pay and performance was a serious
error by these owners. Such retainers weakened principal-agent
incentive mechanisms causing moral hazard behaviour and poorer
performance. The fact that such payments have largely been
abandoned is itself evidence that such non-contingent payments
did not deliver good performance.
We have shown that both the 'traditional pursuits' and 'the
study of horse racing' can be fruitfully combined to yield insights
on the design of payment systems and their performance outcomes.
Paying large sums with insufficient performance incentives built in
is inferior to carefully constructed incentive contracts — in the
© Fondazione Giacomo Brodolini and Blackwell Publishers Ltd 1999.
406 Sue Fernie - David Metcalf

spirit of the old song — 'it ain't what you pay it's the way that you
pay it, and that's what gets results'.

Appendix A. Assessing jockeys' performance

Collateral form ratings


Racing Research calculates form ratings for each horse after
each race. The method is fully described in Whitley (1992, 1995).
After a race the collateral form rating is based on:
(a) direct comparisons: comparisons of a horse against those it
has run against in the race;
(b) indirect comparisons: how these opponents have run against
other horses in different races.
After each new race result Racing Research assigns a level to the
race, i.e. rates the winner. This determines the rating for each horse
in the race. The winner's rating — the level for the race — is based
on knowledge of runners' previous performances.
Other horses' form ratings for that race then follow, based on:
(i) weights carried; (ii) distance beaten; (iii) immaturity, i.e. weight
for age allowance. For 2-year-old races Racing Research may also
consider race time (but only around one race in five is truly run).
As a consequence of ratings derived for the current race Racing
Research then revise their assessment of previous races, i.e. a ripple
effect that spreads through many races that season (they are the
only handicapping organization that do this).

Assessing jockeys
Ratings therefore exist for every single horse/jockey perfor-
mance, e.g. over 40,000 for the flat in 1994. Whitley (1992)
describes how these ratings are used to get performance indicators
for jockeys:
we could look at how each horse had performed, in terms of our form
ratings, under its different riders. There was a huge amount of data for
the busiest riders so that variations due to distance, going and so on
would virtually cancel out. From all this data, figures for each rider
could be derived by a straightforward mathematical process. The
initial concept is simple and the meaning of the final figure is clear. It is
a measure of how horses have run, in terms of form ratings, for the
© Fondazione Giacomo Brodolini and Blackweli Publishers Ltd 1999.
Jockeys' Pay and Performance 407

rider relative to other riders — no more, no less ...This method


exposes what the rider is really contributing to his mount's
performance...we derive jockey ratings by effectively eliminating the
ability of the horse from the performance.
Further, in correspondence, John Whitley reports that he can, in
fact, control for variations due to distance, going, etc. but 'their
inclusion proves practically irrelevant so the factors do evidently
cancel out'.
The Racing Research method is:

Stage i
For each jockey (f) calculate (each year)

where h is the horses ridden, k is each other jockey who has ridden
horse h, y is the average rating for the horse (thus, for example, y^/,
is the average rating for horse h when ridden by jockey y).
We can think of each jockey having his own table (see Table Al).

Table Al
^ ^ ^ ^ ^ Jockey Jockey 7 average rating (yy^) minus separate
Horse ^ ^ ^ ^ ^ average rating by each other jockey (y^/,) X] jockeys

1
2
3
etc
2 horses X

Of course, many cells will be blank because no comparisons are


possible. It is the X in the south east corner that we are interested
in for each jockey.

Stage ii
(a) convert the X scores to pounds;
(b) Calibrate the resulting figures such that the performance
figures have a mean of 10 Ib.
© Fondazione Giacomo Brodolini and Blackwell Publishers Ltd 1999.
408 Sue Fernie - David Metcalf

Appendix B

Table Bl. Current pay (£) regressed on current Racing Research


performance, 1988-95. Ordinary least squares
Sample, pay definition Regression coefficient Constant

Total pay
Unbalanced panel, levels 10,928 -63,596 0.22
(10.3) (-5.3)
Unbalanced panel, logs 1.69 6.63 0.30
(12.1) (20.2)
Balanced panel, levels 12,644 -71,740 0.22
(6.3) (-3.2)
Balanced panel, logs 2.03 6.01 0.33
(8.2) (10.4)
Riding fees only
Unbalanced panel, levels 2,358 -3,318 0.29
(9.7) (-1.2)
Unbalanced panel, logs 0.83 7.95 0.32
(10.0) (40.5)
Balanced panel, levels 2,041 1,101 0.35
(5.3) (0.3)
Balanced panel, logs 0.73 8.26 0.38
(5.7) (27.5)
Prize money earnings only
Unbalanced panel, levels 8,586 -60,454 0.19
(9.6) (-6.0)
Unbalanced panel, logs 2.80 3.17 0.31
(13.2) (6.4)
Balanced panel, levels 10,603 -72,841 0.18
(6.1) (-3.7)
Balanced panel, logs 3.71 1.34 0.37
(9.7) (1.5)

Notes: (i) All regressions are OLS and have Racing Research on the right-hand side and
pay on the left-hand side.
(ii) Observations: unbalanced panel 413, balanced panel 184.
(iii) ^-statistics are in parentheses.
(iv) Time dummies included (to control for increase in all nominal earnings),
(v) Defmition of total pay is riding fees plus prize money,
(vi) Performance measure is Racing Research indicator; see Appendix A.

) Fondazione Giacomo Brodolini and Blackwell Publishers Ltd 1999.


Jockeys' Pay and Performance 409

Appendix C

Table Cl Current pay (£) regressed on current Racing Research


performance, 1988-95. Fixed effects model

Regression Hausman test


Sample, pay definition coefficient Constant (p values)

Total pay
Unbalanced panel, levels 2,873 15,874 0.12 12.85
(2.1) (1.2) (0.12)
Unbalanced panel, logs 0.60 9.11 0.23 11.79
(3.6) (23.7) (0.16)
Balanced panel, levels 3,409 17,807 0.15 3.84
(1.6) (0.7) (0.87)
Balanced panel, logs 0.75 8.97 0.31 5.72
(2.6) (13.2) (0.68)
Riding fees only
Unbalanced panel, levels 1,050 9,359 0.32 11.66
(2.8) (2.5) (0.17)
Unbalanced panel, logs 0.51 8.69 0.38 4.99
(4.2) (31.5) (0.76)
Balanced panel, levels 1,016 11,607 0.47 1.43
(1.8) (1.9) (0.99)
Balanced panel, logs 0.51 8.77 0.53 0.58
(2.7) (20.3) (0.99)
Prize money earnings only
Unbalanced panel, levels 1,823 6,516 0.06 12.43
(1.6) (0.58) (0.13)
Unbalanced panel, logs 0.75 7.88 0.12 19.18
(3.0) (14.0) (0.01)
Balanced panel, levels 2,893 6,200 0.06 3.88
(1.4) (0.3) (0.87)
Balanced panel, logs 1.01 7.57 0.17 13.04
(2.3) (7.4) (0.11)

Notes: See notes to Table 2. The fixed effects model is estimated using individual jockey
dummies. The Hausman test compares the random effects model (Table 2) with the
fixed effects model and indicates that the random effects model is appropriate.

© Fondazione Giacomo Brodolini and Blackwell Publishers Ltd 1999.


410 Sue Fernie - David Metcalf

Notes

' Examples of labour market papers which do test how agents respotid to
different incentives (rather than merely estimating the strength of any relation-
ship) are sparse, but see, for example, Lazear (1996) on the impact of substituting
piece rates for time rates for workers installing replacement car windscreens, and
Ehrenberg and Bognanno (1993) on the prize money-performance relationship in
golf tournaments.
^ Some jockeys in the sample are retained by a particular trainer to ride the
horses from that stable. This strengthens the incentive contract, buttressing it by
what Milgrom and Roberts call 'organizational arrangements' — elements of
vertical integration. In some instances the retainer simply guarantees the rides. In
other cases the jockey receives an extra 1-2 percentage points in prize money
deducted from the owner's share. Such an arrangement represents a more
powerful incentive contract but it proved impossible to get any consistent details
of such fmancial arrangements because they are a private matter negotiated
between the owner, trainer and jockey concerned.
^ We investigated the links between reputation and performance in other ways
too. Racing Research performance was regressed in turn against wins, rides, strike
rate, total prize money won and returns to level stakes separately for this year,
next year and 2 years' time (i.e. Racing Research performance was entered
currently, then lagged 1 year, then lagged 2 years). In each case the regression
coefficient and the R^ get larger over time: Racing Research performance is more
strongly associated with the other (less good) performance indicators in 2 years'
time than it is with current performance. True performance takes a couple of
years to get fully recognized. Reputation — the prerequisite for the quantity and
quality of rides — lags performance both on the way up and on the way down.
'' It has been suggested to us that our characterization of the retainer system is
incorrect and that the system might have very high powered incentives, i.e. the
pay is high and some jockeys lose their retainers. But the evidence suggests that
the majority performed worse when retained than not retained and once retained
kept their retainer until the whole system crumbled in 1993-94.

References

Baker, G., Gibbons, R. and Murphy, K. (1994) 'Subjective Performance


Measures in Optimal Incentive Contracts', Quarterly Journal of Economics
109: 1125-1156.
Beyer, A. (1983) The Winning Horseplayer, New York: Houghton Mifflin.
British Horseracing Board and Jockey Club (1993) Rules of Racing and
Instructions.
Brown, W. (1962) Piecework Abandoned, London: Heinemann.
Conyon, M., Gregg, P. and Machin, S. (1995) 'Taking Care of Business:
Executive Compensation in the UK', Economic Journal 105, May: 704-
714.
Ehrenberg, R. and Bognanno, M. (1990) 'The Incentive Effects of Tournaments
Revisited: Evidence from the European PGA Tour', Industrial and Labor
Relations Review 43 (3), February: 74-88.

© Fondazione Giacomo Brodolini and Blackwell Publishers Ltd 1999.


Jockeys' Pay and Performance 411

Flanders, A. (1970) 'Pay as an Incentive' in Management and Unions, London:


Faber: 72-82.
Green, W. (1991) Econometric Analysis, New York: Macmillan.
Gregg, P., Machin, S. and Szymanski, S. (1993) 'The Disappearing Relationship
between Directors' Pay and Corporate Performance', British Journal of
Industrial Relations 31 (1), March: 1-10.
Hall, B. and Liebman, J. (1997) 'Are CEOs Really Paid Like Bureaucrats?',
NBER Working Paper 6213, October.
Home Affairs Committee (1991) Levy on Horserace Betting, HC 146 I, II, May.
Jensen, M. and Murphy, K. (1990) 'Performance Pay and Top Management
Incentives', Journal of Political Economy 98(2), April: 225-264.
Jockey Club and Horse Racing Advisory Council (1991) The Needs of Racing: A
Eair Price for Racing's Product, September.
Lazear, E. (1995) Personnel Economics, Cambridge, MA: MIT Press.
Lazear, E. (1996) 'Performance Pay and Productivity', Stanford University,
mimeo. May.
Main, B. (1995) 'The Governance of Remuneration for Senior Executives', Paper
presented at Conference on Structure of Corporate Governance, LSE, July.
Main, B., Bruce, A. and Buck, T. (1996) 'Total Board Remuneration and
Company Performance', Economic Journal 106(439), November: 1627-
1644.
Milgrom, P. and Roberts, J. (1992) Economics, Organization and Management,
Englewood Cliffs, NJ: Prentice Hall.
Millan, J. (1995) 'Running on'. Odds On, 36, February.
Montgomery, S. (1995) 'The Breeding Season', Inside Racing, January.
Munting, R. (1987) Hedges and Hurdles: A Social and Economic History of
National Hunt Racing, London: J. A. Allen.
Pacemaker and Thoroughbred Breeder (1995) Sires for '95, London: Haymarket
Publishing.
Richardson, R. and Wood, S. (1989) 'Productivity Changes in the Coal Industry
and the New Industrial Relations', British Journal of Industrial Relations
27(1), March: 33-54.
Roberts, M. and Tanner, M. (1994) Michael Roberts: A Champion's Story,
London: Headline.
Roy, D. (1952) 'Quota Restrictions and Goldbricking in a Machine Shop',
American Journal of Sociology 67 (2), 427-442.
Simon, H. (1991) 'Organisations and Markets', Journal of Economic Perspectives
5(2), Spring: 25-44.
Stiglitz, J. (1991) 'Symposium on Organisations and Economies', Journal of
Economic Perspectives', 5(2), Spring: 15-24.
Whitley, J. (1992) 'Assessing Jockeys', Computer Racing Eorm, Brighouse: Racing
Research.
Whitley, J. (1995) 'A Small Flat World', Computer Racing Eorm, Brighouse:
Racing Research.

1 Fondazione Giacomo Brodolini and Blackwell Pubiishers Ltd 1999.

You might also like