Omega 35 (2007) 202–210www.elsevier.com/locate/omega

Motor vehicle recalls: Trends, patterns and emerging issues

Hilary Batesa, Matthias Holwegb, Michael Lewisc, Nick Oliverd,∗aWarwick Business School, University of Warwick, Coventry, UK

bCentre for Competitiveness and Innovation, Judge Institute of Management, University of Cambridge, Trumpington Street,Cambridge CB2 1AG, UK

cOperations and Supply Group, School of Management, University of Bath, Bath BA2 7AY, UKdCentre for Competitiveness and Innovation, Judge Institute of Management, University of Cambridge, Trumpington Street,

Cambridge CB2 1AG, UK

Received 2 November 2004; accepted 20 May 2005Available online 19 July 2005

Abstract

This paper examines patterns and trends in motor vehicle safety recalls using a dataset based on 23.1 million vehiclesregistered in the UK between 1992 and 2002. A safety recall occurs when vehicle manufacturers call vehicles that have beensold and are in use back to their dealerships for safety-related remedial work. Safety recalls can be a strategic concern forcar makers, having the potential to damage brand value, reduce stock price as well as resulting in significant direct costs. Thedata from this study show that the incidence of vehicle recalls is increasing—between 1998 and 2002 there was an average ofover 120 recall incidents per annum in the UK, compared to less than 50 per annum between 1992 and 1994. Total numbersof vehicles recalled show no clear trend over time, but the absolute level of recalls year on year is very high: in the UK, 10.8million vehicles were recalled during 1992–2002, representing 47% of all vehicle UK registrations in the period. Moreover,there are substantial differences in recall rates between different car manufacturers, suggesting that recall rates may be a usefulindicator of process performance in the automotive design-and-production value chain. European and American producershave recall rates that are nearly three times greater than their East Asian counterparts. This paper concludes with some ideasthat may explain these patterns and suggests an agenda for further research.� 2005 Elsevier Ltd. All rights reserved.

Keywords: Automobile industry; Reliability; Manufacturing; Management

1. Introduction

The automotive industry faces many challenges. It suf-fers from global long-term assembly overcapacity, which isdepressing profitability. In 2000, the installed global assem-bly capacity for motor vehicle production was 79.2 millionunits (measured on the basis of two shifts per day in the

∗ Corresponding author. Tel.: +44 1223 338179;fax: +44 1223 339701.

E-mail addresses: Hilary.Bates@wbs.ac.uk (H. Bates),m.holweg@jims.cam.ac.uk (M. Holweg), m.a.lewis@bath.ac.uk(M. Lewis), n.oliver@jims.cam.ac.uk (N. Oliver).

0305-0483/$ - see front matter � 2005 Elsevier Ltd. All rights reserved.doi:10.1016/j.omega.2005.05.006

assembly plants), compared to global sales of 58.8 millionunits—an overcapacity of 35% [1]. Ellen Hughes-Cromwik,the chief economist of Ford Motor Company, estimatedthat the North American industry alone had an overcapacityof 3.3 million units [2]. Moreover, the problem is not con-fined to established markets: in China, capacity is predictedto rise to 6.4 million units by 2007, compared to a localdemand forecast of 3.1 million units [3]. A spate of merg-ers and acquisitions in the industry amongst both vehiclemanufacturers and component makers has been one conse-quence of this. Further rationalization amongst incumbentsin the developed world seems inevitable, especially as newentrants in the developing world continue to create new

H. Bates et al. / Omega 35 (2007) 202–210 203

capacity. In such an environment, anything that risks dam-age to brand value or drives up operating costs is obviouslyhighly undesirable—hence, recalls are of strategic impor-tance to the car makers.

Safety recalls occur when car manufacturers identify post-production safety concerns with their vehicles. When thishappens the car makers will try to locate all affected vehi-cles in order to have faults rectified, at no cost to the owners,at franchised garages. Vehicle safety recalls are, by defini-tion, significant events: they are carried out in the name ofprotecting drivers, passengers and the general public but forvehicle manufacturers (and component suppliers) they canbe expensive and adversely impact brand value. In Novem-ber 2003 for instance, a UK tabloid newspaper reported arecall by French car maker Peugeot-Citroen with the head-line “Brake Scare on 270,000 Motors” and opened the ar-ticle with the words “FRENCH car giant Peugeot is recall-ing 270,000 models because the brake pedals could FALLOFF” (The Sun, 21 November 2003).

Some of the most dramatic examples of the dynamics andconsequences of vehicle recalls to date have been seen inthe US. For example, during the 1980s, it was alleged thatthe Audi 5000 model (sold as the Audi 100/200 series inEurope) was prone to “sudden acceleration” at low speed.The company denied that there was a problem, but the ef-fect on sales was dramatic. With nearly 1800 claims of un-intended acceleration incidents in the 1980s, Audi becamesynonymous with the term “sudden acceleration.” Accord-ing to the Centre for Auto Safety, despite several deaths,dozens of injuries, five related recalls and a Swedish De-fence Agency study showing that cruise control malfunctionscould cause sudden acceleration, Audi claimed that “pedalmisapplication” caused the incidents but lost 80% of its USmarket share in the US as a consequence of the episode [4].One of the most significant cases of recent years was an al-leged problem with tyres on the Ford Explorer, in which itwas claimed that Explorers were prone to roll following atyre blow out. Firestone recalled 14 million tyres in 2000and analysts estimate that this particular recall cost about$750 million—not counting any liability associated with anylawsuits. This incident led to a dispute between Ford andFirestone, the manufacturer of the tyres. The opening linesof an article about the Ford/Firestone incident in the Wash-ington Post demonstrate the often emotive nature of reportsconcerning vehicle safety issues:

‘From the beginning of the Firestone tire recall, FordMotor Co. officials have insisted that the accidents thatkilled 101 Americans, most of them in Ford Explor-ers, are a Firestone tire problem’ (Washington Post, 9October 2000). [Our emphasis].

2. Previous studies

Given the high profile of vehicle recalls, what researchinto recalls has already been conducted? Most existing stud-

ies, all US-based, have focused on the influence and natureof regulation [5–7] and the external and indirect costs as-sociated with consumer, capital market and media reactionsto vehicle recalls [7,8].

The largest category of literature on recalls considers theirimpact on market share and share price, although the resultsof these studies are not consistent. Early studies of marketreaction to recalls [9,10] found all but the most severe recallshad limited influence on customer demand for new vehicles[11]. However, work on the impact of recalls on used carvalues found that recalls had a substantial negative impacton the residual values of recall-prone models [12].

Studies of the relationship between recalls and stock(share) price indicate that the damage to stock pricescaused by recalls was actually greater than the directcosts of recalls [13]. However, this conclusion has beencriticized [14] and challenged [6] on the grounds that onan event-by-event basis 40% of reactions to recalls werepositive. Other commentators have re-asserted the predom-inance of the indirect costs of recalls, especially for certaintypes of recall [15]. Recalls affecting components such asairbags, for example, produce a more negative effect thanthose that affect other parts of the car, such as the heatingsystem [16].

Given that recalls are public events, the influence of mediareporting has also been investigated. For instance, the cover-age of recalls affecting vehicles produced by Ford, GM andChrysler in Wall Street Journal has been shown to have neg-ative impact on the stock price of these companies—muchmore so, in fact, than official notices by the car makersthemselves [17]. Of course, media coverage of recalls tendsto be driven by the size, severity and total costs (includ-ing punitive legal awards) of the recall [18,19], so it maybe these factors, rather than coverage per se, that is drivingany effect on sales. Interestingly, many of the factors asso-ciated with the success of a recall, such as high responseand repair rates, are also those associated with strong neg-ative consumer reaction to the recall, such as the recencywith which the recalled model has been launched, and theseverity—and hence newsworthiness—of the hazard posedby the fault [7,20].

As will be obvious from the review thus far, the existingliterature on recalls has largely focused on the economiceffects of recalls, rather than issues such as their underly-ing causes, although Krueger and Mas’ analysis of the FordExplorer/Firestone tyre recall is an exception to this [21].There has been no work on propensity to recall across differ-ent manufacturers or analyses of trends in recall rates overtime. There is a very limited literature on operational issuessuch as the direct costs of recalls, or on the organizationalor other failures of which a recall is the ultimate outcome.An exception is Dardis and Zent’s analysis of Ford’s in-famous strategy for dealing with a problem with the 1978Pinto, in which Ford concluded that the costs of the de-sign change were likely to be greater than the costs of lawsuits against the firm and therefore eschewed any design

204 H. Bates et al. / Omega 35 (2007) 202–210

modifications, a conclusion that ultimately proved to be verycost-ineffective [22].

The review also reveals that vehicle recalls have receivedscant attention in Operations Management (OM) research.This is particularly surprising when noting the strong res-onance between recall issues and what have historicallybeen core OM concerns, such as how different firms andproduct development and manufacturing systems performwith respect to product quality. There is some evidencethat Japanese manufacturers have fewer recalls than theirUS counterparts [15], a picture consistent with compara-tive studies of manufacturing productivity and quality con-ducted under the auspices of global research programmessuch as the International Motor Vehicle Programme (IMVP)[23,24]. However, it is important to note that although re-calls may appear superficially similar to other measures ofdefect, there are also important differences between recallsand the quality measures employed by the IMVP research.

The IMVP quality data are based on J.D. Power data de-rived from surveys of new vehicle buyers, identifying defectsthat relate to the vehicle assembly process. These surveyscover defects noted by owners in the first few months ofownership of the vehicle—typical examples might be paintblemishes, squeaks and rattles, equipment malfunctions andso on. Recalls often relate to faults that are invisible to thevast majority of owners—the precautionary replacement ofa pin on a brake pedal, for example, is not something thatthe J.D. Power survey would have detected unless the parthad actually failed. Of course some of the underlying capa-bilities that deliver high quality as measured by J.D. Power-type surveys may well be those that minimize the risk oferrors that subsequently show up as recalls.

In this paper, recalls are treated as an additional measureof performance of the automotive design, supply and pro-duction chain, and moreover, one that has so far received lit-tle attention, and—as will be shown below—seems to showworsening performance over time, contrary to the improve-ments that have been according to more traditional measuresof vehicle quality over the last decade.

This paper therefore sets out to address three mainquestions.

1. Given the often high-profile nature of vehicle recalls,how prevalent are recalls, particularly when normalizedfor levels of vehicle registrations?

2. Is there any trend over time in numbers of vehicle recalls,and if so, what might be the explanations behind anysuch trend?

3. Are there significant differences in recall rates betweendifferent manufacturers, and if so, what seem to be themost fruitful avenues for research into the explanation ofsuch differences in the future?

Thus, this paper aims to provide baseline data on the scaleof the problem of recalls. It examines the prevalence of ve-hicle recalls, differences between car makers in propensity

to recall, and sets out a future research agenda on the prob-lem of recalls.

3. Research methods

The data presented in this paper comprise two mainelements—data on vehicle recalls and data on vehicleregistrations. The data on recalls were extracted from apublicly accessible database in the UK provided by theVehicle and Operator Services Agency (VOSA). In the UK,car recalls are operated under a code of practice1 agreedbetween VOSA and the carmakers. VOSA operates a web-site (http://www.vosa.gov.uk/vosa/) which contains detailedinformation on vehicle recalls in the UK. This covers man-ufacturers, models, dates of production, details of the faults,and sometimes the Vehicle Identification Numbers (VINs)of the vehicles that are affected. The dataset covers vehiclerecalls by over 40 vehicle manufacturers between 1992 and2002. The data on vehicle registrations were obtained fromJust-Auto, a commercial automotive intelligence company.The registration data are presented on a model-by-modelfor each year. In combination these two sources of datareveal that 23,099,246 new vehicles were registered in theUK between 1992 and 2002. The registrations data cover507 models sold in the UK during this period. Note thatthe figure of 23 million vehicles represents the number ofnew units registered between 1992 and 2002—it is not ameasure of the size of the vehicle parc.

The recall analysis covers 40 vehicle manufacturers, anda total of 493 different models, which account for 99.7% ofthe UK market in 1992–2002. Several small-scale vehiclemanufacturers were excluded from the analysis due to in-consistencies in the data for these firms. These exclusionsincluded TVR, Rolls-Royce, Caterham, Westfield, and Mor-gan. These account for the discrepancy in the number ofmodels between the two datasets.

There are a number of methodological issues in aligningthe recall and registration data, coming as they do from twodifferent datasets. Vehicles recalled may be spread over buildperiods that stretch over several years. The recalls databasedoes not indicate which vehicles were built on which datesand therefore it is necessary to make a proportional split ofvehicles over the build period in order to link it back to theregistration data, which are provided on the basis of calendaryears. In some cases, build dates are not provided at all, andin others several models from the same manufacturer are

1 Although an informal agreement, a manufacturer notifyingVOSA of a safety defect is allowed limited access to normallyconfidential national registration records to enable them to contactthe owners about the problem. VOSA estimate that approximately85% of vehicle recalls are voluntary actions on the part of the carmakers, 15% are initiated by VOSA due to issues identified bydealers, vehicle owners, the Trading Standards Authority, or otherbodies, such as consumer groups.

H. Bates et al. / Omega 35 (2007) 202–210 205

21.0%

32.9%

25.0%

12.0%

8.2%

0.5% 0.4%0%

10%

20%

30%

40%

BuildYear

1 Year 2 Years 3 Years 4 Years 5 Years 5 Years+

% Recall Volume in Years after Build Year

Fig. 1. Delay of recall reports in relation to vehicle build date.Example of Ford, all models from 1992 to 2002.

recalled under the same notice, making it difficult to ascribea precise number of recalled units to each model coveredby that recall.

To complicate matters further, there is a considerable timelag in the recall incidents. Once a safety-relevant issue isdiscovered, a car manufacturer issues a recall. Recalls mayoccur in the build year of the vehicle, or some years afterthe vehicle has been registered. Fig. 1 shows a typical distri-bution of delays between a vehicle being built and recalled.

As can be seen, approximately 50% of safety recalls areinitiated a year or more after the vehicle was built, though80% occur within 3 years of the build date. In the analysis inthis paper, recall incidents were recorded in the year the re-call notification was issued through VOSA. Recall volumes,however, are attributed to the calendar year in which a ve-hicle is built. Thus, some noise in the dataset is inevitable,given that the entire vehicle production of one calendar yearis not sold in the same calendar year. Sales registrations inJanuary of 1 year, for example, will contain vehicles pro-duced in the final months of the previous year.

The longer the period covered by the dataset, the lesslag is a problem, partly because most recalls occur in theearly years of production of a new model, partly becauselong data periods will include the complete life of a modeland therefore all registrations and all recalls attributable tothat model will be captured. The error term is negligiblewhen several years of data are considered, especially givenan average product life cycle of 5–7 years in Europe [24].However, the problem of lag means that annual figures forrecall rates need to be treated with some caution.

4. Research findings

Five main variables were constructed from the dataset.These are:

1. Total volume of vehicle recalls per manufacturer per an-num. This refers to the total number of vehicle recalls in

a given year. This refers to the cumulative total for all ve-hicles for that manufacturer for each calendar year. If thesame vehicle is recalled multiple times during the year,then each of these counts as a separate recall. “Vehiclerecalls” are not therefore synonymous with “number ofvehicles recalled”.

2. Total volume of vehicle recalls per manufacturer per an-num per 1000 registrations. This normalizes the measuredescribed in (1) for numbers of registrations.

3. Total numbers of recall incidents per manufacturer perannum. This refers to the number of separate recall in-cidents per year, independent of how many vehicles ormodels are affected by each incident. Thus, a single recallstemming from a common problem that affects three dif-ferent models would count as one recall incident. Thus,“number of recall incidents” effectively equates to “num-ber of recall announcements”.

4. Number of recall incidents per annum per 100,000 vehi-cles registered. This takes the measure described in (3)and controls for the number of vehicles registered in agiven year. This is important in order to distinguish be-tween any increase in recalls attributable to a general in-crease in numbers of vehicles sold, and real increases inthe frequency of recalls.

5. Number of models offered by each manufacturer. Thisrefers to the total number of models offered over the12-year period. This is used to explore the relationshipbetween product variety and propensity to recall.The results are presented in three parts. First, we examine

trends in recalls over time. Secondly, we examine absolutelevels of recalls and differences in recall rates by differentmanufacturers. Thirdly, we explore how the complexity ofproduct ranges (manifested by numbers of different modelsoffered to the market by each manufacturer) is related topropensity to recall.

4.1. Absolute levels of recalls and trends over time

The VOSA dataset shows that between 1992 and 2002there were 838 separate vehicle safety recall incidents in theUK representing a cumulative total of 10,752,093 vehicle-recalls, an average of 977,463 per year. Combined with thedata on vehicle registrations, this shows that the volume ofvehicle recalls represents 47% of all vehicles registered dur-ing that period. One might conclude from this therefore thatthe average probability of a specific vehicle being subjectto a recall is nearly 50%. However, this is not entirely accu-rate. A significant proportion of vehicles are recalled morethan once, so the probability of an individual vehicle beingsubject to a safety recall is accordingly lower.

It is not possible to derive the exact figure for the numberof individual vehicles recalled from the VOSA database asthe VIN numbers of the vehicles affected by the respectiverecalls are not always recorded. Nevertheless, it is possibleto estimate the interval of the probability that a particularvehicle registered in the period between 1992 and 2002 will

206 H. Bates et al. / Omega 35 (2007) 202–210

43 51 4655

6854

85 73

103

123

137

0

500,000

1,000,000

1,500,000

2,000,000

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

Nu

mb

er o

f Veh

icle

s

0

40

80

120

160

Nu

mb

er o

f In

cid

ents

Total Number of Vehicles Recalled Total Number of Recall Incidents

Fig. 2. UK recall incidents and number of vehicle recalls1992–2002.

be subject to at least one (or more) safety recalls duringthis period. We define Vij as the recall volume of a givenrecall incident i for vehicle manufacturer j , and the overallregistrations for the period for a manufacturer i as Ri . Theupper bound of this interval assumes that all recall incidentsaffect different vehicles, i.e. not a single vehicle is recalledmore than once. However, as vehicles can be recalled morethan once, and since one recall incident can affect morethan one model of the manufacturer in question, the lowerbound of the interval assumes a maximum overlap of therecall incidents. This assumes that all recall incidents at agiven manufacturer are subsets of the single largest recallincident that occurred during that period. As shown in Eq.(1), these assumptions allow for the quantification of thelikelihood that a vehicle registered in the period between1992 and 2002 in the UK will be recalled with a probabilityof at least 12%, with the maximum possible recall volumebeing equal to 47% of all vehicles registered.

Recall propensity, upper and lower bounds:

Pupper =∑n

i=1∑m

j=1 Vij∑n

i=1 Ri= 0.469,

Plower =∑n

i=1Maxmj=1 (Vij )

∑ni=1 Ri

= 0.119. (1)

The VOSA data available unfortunately do not permit fur-ther refinement of the interval. However, the authors con-ducted two interviews with managers from a global vehi-cle manufacturer (one was the manager of European testingand validation, the other the manager of supplier warrantyissues) to obtain estimates of the probability of a single ve-hicle being recalled. The interviewees confirmed the overallhigh levels of recall incidents and volumes, and estimatedthe probability of a specific vehicle being subject to oneor more recalls at around 25%—close to the median of thetheoretical range calculated above.

Fig. 2 shows the number of recall incidents per annumbetween 1992 and 2002, and the total number of vehiclerecalls for the same period. There appears to be an upward

43.137.7

34.4

64.1

36.430.3

35.8

46.1

62.6

74.9

41.0

0

20

40

60

80

100

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002Rec

all V

olu

mes

as

% o

f R

egis

trat

ion

s

Fig. 3. Recalls as a percentage of vehicle registrations 1992–2002.

y = 0.272x - 539.68

R2 = 0.7518

0

20

40

60

80

100

120

140

160

1990 1992 1994 1996 1998 2000 2002 2004

Nu

mb

er o

f In

cid

ents

0.00

1.00

2.00

3.00

4.00

5.00

6.00

Nu

mb

er o

f In

cid

ents

per

100

,000

R

egis

trat

ion

s

UK: Number of Recall Incidents UK: Recalls per 100,000 registrations

Linear (UK: Number of Recall Incidents) Lin. Regression UK Recalls per 100,000 Reg.

Fig. 4. Upward trend on recall incidents (UK).

trend in the absolute volume of vehicle recalls per annumbetween 1992 and 2001, but this falls back to close to thelong run average in 2002. However, when vehicle recallvolumes are normalized for levels of registrations, there isno evidence of any trend, upwards or downwards, over time.This is demonstrated by Fig. 3, which shows numbers ofrecalls as a percentage of vehicle registrations.

Considerable variation in the number of vehicle recallsfrom year to year is also apparent in Figs. 2 and 3, typicallydue to a small number of incidents that affected very largenumbers of vehicles. This is apparent in 1995 and again2000–2001, in which the large peaks in the recall rates weredriven by just three separate recall incidents affecting severalmodels in the GM-Vauxhall range.

However, although there is no trend in recall volumes overtime when these are normalized for registrations, the numberof recall incidents shows a fairly consistent rise over thelast 12 years, showing a two- to threefold increase betweenthe period 1992–1994 and the period 2001–2002. As withvolumes of recalls, the number of recall incidents per annumwere normalized by creating an index of “recall incidentsper 100,000 new vehicle registrations”, thus controlling forany increase in vehicle registrations. The results of this areshown in Fig. 4.

Even when normalized for numbers of annual registra-tions, recall incidents show the same pattern of increas-ing over time, from two to three incidents per 100,000

H. Bates et al. / Omega 35 (2007) 202–210 207

registrations in 1992–1995 to around five in 2001–2002. Re-gression analysis shows a positive, linear annual increase ofrecall incidents in the order of 26%, with a regression co-efficient of R2 = 78%, demonstrating a significant trend ofan increasing incidence of vehicle recalls.

4.2. Recalls rates of different vehicle manufacturers

Vehicle manufacturers vary enormously in terms of boththe volumes of vehicles that they produce and also the va-riety of models that they produce. For example, the averagenumber of registrations per annum over the 11-year periodvary from 125 vehicles for Maserati to 393,538 vehicles forFord, with an average of 52,346 registrations per annumacross all manufacturers. As is to be expected, high-volumemanufacturers with high levels of registrations tend to showhigh volumes of recalls, but there are exceptions to this.Table 1 ranks vehicle manufacturers according to recall vol-umes as percentage of registrations.

There are several points to note from Table 1. First,the range of recall rates over the 11-year period is verylarge—according to the VOSA data, two manufacturers(Isuzu and Subaru) had no recalls at all, but several (Ferrari,BMW-Mini, Land Rover and Jeep) had recall rates closeto, or even above, 100%. Such enormous differences raiseinteresting questions about the reasons behind corporatepropensity to recall.

Second, there is no obvious pattern to the ranking, in termsof what types of manufacturers are prone to recalls. Theposition of Ferrari is perhaps not surprising, as a very low-volume, craft-based producer. However, some very high-volume producers such as Vauxhall (GM) and Volkswagenalso appear in the top 10 companies in terms of recall rates.

Third, consistent with previous findings on manufactur-ing performance in the global auto industry [23], it is strik-ing that eight of the 10 companies with the lowest recallrates have East Asian connections, either in terms of owner-ship, or manufacturing location, or both. Five of these eightare Japanese-owned. If one regards recall rate as anotherindicator of product quality (in the sense that the problemsthat recalls are supposed to solve represent errors of com-mission or omission in design, manufacturing, or compo-nent supply), then these findings are consistent with oth-ers that have demonstrated performance superiority on thepart of Japanese vehicle producers with respect to productquality [23].

The registration data show that the number of models of-fered by each manufacturer over the period of 1992–2002varies considerably across manufacturers, from 30 models(Toyota) to one model (Isuzu), with an average of 12.3 mod-els per manufacturer.

One might expect product variety, as measured by thenumbers of different models produced by a manufacturer tobe linked to propensity to recall. More models mean greaterdesign, manufacturing and supply chain complexity, whichin turn are likely to mean more errors. The link between

Table 1Ranking of manufacturers: UK recall volumes as percentage ofregistrations (1992--2002)

Ranking Manufacturer Recall volumes as %of total registrations

1 FERRARI 1502 BMW-Mini 1103 LAND ROVER 1034 JEEP 955 ASTON MARTIN 906 VAUXHALL 797 LOTUS 788 CHRYSLER 749 VOLKSWAGEN 6910 LANCIA 6611 MASERATI 6312 VOLVO 5313 MITSUBISHI 5214 FORD 5115 FIAT 5116 SAAB 5017 CITROEN 4918 BENTLEY 3919 DAIHATSU 3920 AUDI 3521 PORSCHE 3422 SEAT 3323 MAZDA 3224 PEUGEOT 2825 SKODA 2826 MERCEDES 2527 MG-ROVER 2528 NISSAN 2429 RENAULT 2330 HYUNDAI 2331 TOYOTA 2132 ALFA ROMEO 2133 DAEWOO 2034 HONDA 2035 JAGUAR 1936 SUZUKI 1237 PROTON 1038 KIA 539 ISUZU 040 SUBARU 0

Average 47%

product variety and recall propensity is explored in the cor-relation matrix shown in Table 2.

Number of models offered is used as a measure of theproduct variety offered by manufacturers over the period inquestion. Table 2 shows that the number of recall incidentsstrongly correlates with the number of models offered, andthat the number of recall incidents correlates with the volumeof vehicles recalled. However, once there is a correction forthe level of vehicle registrations, the correlations between

208 H. Bates et al. / Omega 35 (2007) 202–210

Table 2Product variety and propensity to recall

Pearson correlation Number of Recall volume Number of recall Index: recall Index: recall(2-tailed) N = 40 models offered incidents incidents per 100,000 volume per 1000

1992–2002 registrations registrations

Number of models 1.00offered 1992–2002 —

Recall volume .28 1.00p = .07 —

Number of recall .47** .62** 1.00incidents p = .00 p = .00 —

Index: recall incidents −.287 −.18 −.13 1.00per 100,000 registrations p = .072 P = .27 P = .43 —

Index: recall volume −.311* .23 .12 .58** 1.00per 1000 registrations p = .050 p = .16 p = .45 p = .00 —

∗Correlation significant at the 0.05 level (2-tailed).∗∗Correlation significant at the 0.01 level (2-tailed).

Models offered 1992-2002

403020100

Inde

x: r

ecal

l vol

ume

div

by r

egis

trat

ions

1.6

1.4

1.2

1.0

0.8

0.6

0.4

0.2

0.0

-0.2

VOLVO

VOLKSWAG

VAUXHALL

TOYOTA

SUZUKI

SUBARU

SKODASEAT

SAAB

RENAULT

PROTON

PORSCHE

NISSANMAZDA

MASERATI

LOTUS

LAND ROV

LANCIA

KIA

JEEP

JAGUAR

ISUZU

HONDA

FORD

FERRARI

DAIHATSU

CITROEN

CHRYSLER

BMW-MINI

AUDI

ASTON MA

Fig. 5. Recall volumes per registration by number of models offered(linear regression: adj. r2 = .073, with t-value of 2.02).

product variety and recall incidents and recall volumes dropto borderline levels of statistical significance (−.29, p =.07 and −.31, p = .05 respectively). This implies that theability of manufacturers to handle product variety variessignificantly; some may be capable of handling high productvariety without incurring recalls, whereas others are not.

To further explore this, the position of each manufactureris shown in Fig. 5 as a scatter plot of “recall volumes as apercentage of registrations” and “number of models”.

The companies below the line show fewer recalls relativeto the number of models offered (the proxy for product va-riety) than those above the line. The further to the right in

terms of its position on the scatter plot, the more productvariety a company has to handle. Toyota, for example, hasthe highest product variety of all but is near the bottom ofthe field (31 out of 40) in terms of volume of recalled vehi-cles as a percentage of registrations. This suggests an abilityon Toyota’s part to simultaneously offer relatively high va-riety whilst retaining tight control of product developmentand manufacturing processes. At the opposite extreme areFerrari, Landrover and Jeep who despite having low modelvariety still suffer from very high numbers of recalls.

5. Discussion and conclusions

Recalling a vehicle can be a costly business for car mak-ers, suppliers and dealers, both financially and in terms ofreputational damage. More than one million vehicles havebeen recalled each year in the past 5 years in the UK, andit is estimated that over 30 million vehicles have been re-called in the UK since 1979. Consequently, it is surprisingthat so little attention has been devoted to the underlyingcauses of vehicle recalls. Some research has been reportedinto the impact of recalls on share prices and demand, butlittle is known about the factors that drive levels of recalls.Overall, we estimate that one in four vehicles in the UKsuffers at least one recall during its life, and some will berecalled a number of times. Moreover, an upward trend insafety-relevant recall incidents is apparent in the UK. Thenumber of vehicles affected as a percentage of registrationsis, on average, fairly stable.

H. Bates et al. / Omega 35 (2007) 202–210 209

A detailed analysis of the reasons behind these patternsis beyond the scope of this paper, but we propose that thefollowing areas deserve research attention in the future:

1. Why are recall incidents increasing over time? Thereare number of possible explanations of this, but thesefall into essentially two categories—those that relate tothe vehicles themselves (i.e. an increase in safety-relatedproblems) and those that relate to the environment (e.g.changes in the threshold for reporting such problems).For example, it may be that manufacturers are more fear-ful of litigation than they used to be, and that this hasincreased their propensity to recall, even though thereare no more faults with their vehicles now than was thecase 12 years ago. On the other hand, it is well docu-mented that product development lead times in the au-tomotive industry have been falling, and are set to fallfurther [25,26]. Is one of the side effects of faster times-to-market that the vehicles have more faults when theyget to market?

2. What are the reasons behind the substantial differences inrecall rates between manufacturers? As Table 1 demon-strates, there are enormous differences between car mak-ers in their propensity to recall, and there appears tobe a geographical effect. East Asian producers compriseone-third of the dataset, but represent eight out of the10 companies with the lowest recall rates. Why is this?Does corporate strategy play a role in propensity to re-call? For example, do car manufacturers who emphasizesafety considerations in their advertising and brand val-ues have a lower threshold to issue recall notices? Dosuch priorities affect design and production processes inways that reduce the likelihood of over-sights that laterresult in recalls?

3. How do recall rates relate to product development pro-cesses and lead times? The importance of this questionis implied by (1) and (2) above. Do companies that haveshort product development lead times show higher recallrates? Is a high recall rate a consequence of curtailedproblem-solving during the development cycle? How isthat some manufacturers, such as Toyota, appear to beable to simultaneously enjoy low recall rates, high prod-uct variety and short product development lead times,whereas others cannot? What capabilities underlie this?How are these developed and sustained, or not, as thecase may be?

4. How do recall rates relate to other developments in theautomotive industry? In efforts to drive down costs, mostautomotive manufacturers are attempting to squeezemore use out of a reduced number of platforms [27]. Thisincreases the complexity of coordination across teamsworking on different models, and also means that a givenplatform has to bear an increased number of model con-figurations. Are increased recalls a consequence of this?Similarly, more and more product development (andmanufacturing) is moving out to suppliers systems sup-

pliers, who themselves are having to develop new skillsas systems integrators. As suppliers become responsiblefor the development and manufacture of an increasingproportion of the car, their role in recall levels is likelytoincrease accordingly. Are increased recalls linked tothis trend?

5. How do recall rates relate to other measures of productquality? This paper has assumed that recall rate does rep-resent a meaningful measure of corporate performance,that in turn can be understood by examination of environ-mental conditions and organizational and product factors.How do recall rates relate to other measures of qualityand efficiency, such as development and manufacturingefficiency and other measures of quality and customersatisfaction?To conclude, further research is needed into product

recalls, and we have suggested some directions that thiscould take. The continued growth of recall incidents is ofstrategic significance to the automakers, and more seri-ously, also implies risk to the public in terms of vehiclesafety. Here, the differences between manufacturers inpropensity to recall need to be further investigated, takinginto account the differences in model range, productionvolumes, and manufacturing and design practices, in or-der to establish the root causes of this intriguing andpotentially worrying phenomenon.

Acknowledgements

The authors are grateful to the ESRC Centre for BusinessResearch at the University of Cambridge for their financialsupport for the work on which this paper is based.

References

[1] Pemberton M. Overcapacity—myth or reality? Newsletter,Autelligence, London, March, 2005.

[2] Chappell L. Analysis: excess capacity will haunt N.A.Automotive News, January 24, 2005.

[3] Connelly M. Overcapacity is a concern. Automotive News,June 14, 2004.

[4] Centre for Auto Safety (2004). www.autosafety.org.[5] Tobin RJ. Safety-related defects in motor vehicles and

the evaluation of self-regulation. Policy Studies Review1982;1(3):532–60.

[6] Bromiley P, Marcus AA. The deterrent to dubious corporatebehaviour: profitability, probability and safety recalls.Strategic Management Journal 1989;10(3):233–50.

[7] Rupp N, Taylor C. Who initiates recalls and who cares?Evidence from the automotive industry. Journal of IndustrialEconomics 2002;50(2):123–50.

[8] Koretz G. How recalls shake stocks. Business Week, January13, 1997. p. 30.

[9] Wynne JA, Hoffer G. Auto recalls: do they affect marketshare?. Applied Economics 1976;8:157–63.

210 H. Bates et al. / Omega 35 (2007) 202–210

[10] Crafton S, Hoffer GE, Reilly RJ. Testing the impact ofrecalls on the demand for automobiles. Economic Inquiry1981;19(4):694–703.

[11] Reilly RJ, Hoffer GE. Will retarding the information flow onautomobile recalls affect consumer demand. Economic Inquiry1983;21:444–7.

[12] Hartman RS. Product quality and efficiency: the effect ofproduct recalls on resale prices and firm valuation. The Reviewof Economics and Statistics 1987;69(2):367–72.

[13] Jarell G, Peltzman S. The impact of product recalls on thewealth of sellers. Journal of Political Economy 1985;93(3):512–36.

[14] Hoffer GE, Pruitt SW, Reilly RJ. The impact of product recallson the wealth of sellers: a re-examination. The Journal ofPolitical Economy 1988;96(3):663–70.

[15] Barber BM, Darrough MN. Product reliability and firmvalue: the experience of American and Japanese automakers,1973–1992. Journal of Political Economy 1996;104(5):1084–99.

[16] Rupp N. The attributes of a costly recall: evidence fromthe automotive industry. Review of Industrial Organization,Forthcoming.

[17] Hoffer GE, Pruitt SW, Reilly RJ. Automobile recallsand information efficiency. The Financial Review 1987;22:433–42.

[18] Garber S, Bower AG. Newspaper coverage of automotiveliability verdicts. Law & Society Review 1999;33(1):93–122.

[19] Rupp N. Newspaper coverage of automotive safety recalls.East Carolina University Department of Economics workingpaper, 2001.

[20] Hoffer GE, Pruitt SW, Reilly RJ. When recalls matter: factorsaffecting owner response to automotive recalls. Journal ofConsumer Affairs 1994; Summer: 96–106.

[21] Krueger AB, Mas A. Strikes, scabs, and tread separations:labor strife and the production of defective Bridgestone/Firestone tires. Journal of Political Economy 2004;112(2):253–89.

[22] Dardis R, Zent C. The economics of the Pinto recall. Journalof Consumer Affairs 1982;16:261–77.

[23] Womack JP, Jones DT, Roos D. The machine that changedthe world. New York: Rawson Associates; 1990.

[24] Holweg M, Pil F. The second century: reconnecting customerand value chain through build-to-order. Cambridge, MA: MITPress; 2004.

[25] Fujimoto T. Shortening lead time through early problemsolving—a new round of capability building in the autoindustry. In: Jurgens U, editor. New product development andproduction networks. Berlin: Springer; 2000.

[26] Oliver N, Primost D. Recent product developmentperformance in the Japanese and UK auto industries. EurOMAconference, Como, 16–18 June 2003.

[27] Pil F, Holweg M. Linking product variety to order fulfilmentstrategies. Interfaces 2004;34(5):394–403.