Oxford Economic Papers

Oxford Economic Papers 2007 59(Supplement 1):i31-i48; doi:10.1093/oep/gpm029

This Article Abstract FREE Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Alert me to new issues of the journal Add to My Personal Archive Download to citation manager Request Permissions Google Scholar Articles by Offer, A. Search for Related Content Social Bookmarking          What's this?

© Oxford University Press 2007 All rights reserved

The markup for lemons: quality and uncertainty in American and British used-car markets c. 1953–73

Avner Offer

All Souls College, Oxford OX1 4AL; email: avner.offer{at}all-souls.ox.ac.uk

	Abstract

TOP Abstract 1 Introduction 2 Argument 3 Data 4 Depreciation 5 Dealer markups 6 The influence of... 7 Conclusion Supplementary material Acknowledgements References

 
Automobile depreciation rates and dealer markups in the United States and Britain during the 1950s and 1960s provide evidence on the effect of asymmetric information on market structures. Initial depreciation was not exceptional, and trade was not disabled. The risk of asymmetric information was not large, and was largely covered by dealer warranties. Adverse selection kicked in as cars aged: markups increased and fixed selling costs caused dealers to withdraw from trading older cars. Despite their lower quality, British makes depreciated less, probably due to different novelty signals and longer styling cycles.

Key Words: JEL classifications: D82 • L15 • L62

	1 Introduction

TOP Abstract 1 Introduction 2 Argument 3 Data 4 Depreciation 5 Dealer markups 6 The influence of... 7 Conclusion Supplementary material Acknowledgements References

 
‘From time to time one hears either mention of or surprise at the large price difference between new cars and those which have just left the showroom.’ (Akerlof, 1970, p.489). Why do cars lose so much of their value as soon as they are sold? Akerlof's explanation is well known—an almost-new car offered for sale is likely to be a ‘lemon’, and is discounted accordingly: ‘most cars traded will be the ‘lemons’, and good cars may not be traded at all. The ‘bad’ cars tend to drive out the good’ (Akerlof, 1970, p.489). This has entered the conventional wisdom: a popularizing economics book has as its sub-title ‘Why you can never buy a decent used car!’ (Harford, 2005).¹

The argument is intrinsically compelling, although originally no evidence was provided to support it. It is not easy to test. It is consistent with three different outcomes: no trading, exceptional discounting, and trading facilitated by dealer warranties (Akerlof, 1970, p.489, 499). Indeed, the insight does not depend for its validity on the realities of the used car market, which only serve as a plausible metaphor (Sugden, 2000). Nevertheless, it is useful to investigate reality, if only as a reminder that it does not always support compelling models. In this paper, some evidence of used-car markets is examined to understand how they really worked.

Several empirical studies have attempted to test the argument in the automobile market (Bond, 1982, 1984; Lacko, 1986; Genesove, 1993; Porter and Sattler, 1999; Emons and Sheldon, 2002). All of them have concentrated on adverse selection in vehicles that are typically more than a year old, and mostly older than that. Overall, they have found the evidence of adverse selection to be absent, or at best, weak.² The existence of a large private market in second-hand cars also suggests that the risk of ‘lemons’ was not in itself sufficient to inhibit trade seriously in cars at any age.

However, these are not tests of Akerlof's own example. Nobody expects an older used car to be ‘as good as new’. Some wear and tear is expected, and it is not obvious that post-purchase repairs arise from ‘pre-existing conditions’ which were known to sellers in advance. The tests were typically rather weak ones, which categorized lemons as cars that required more repairs (or that were sold more frequently), and attempted to identify some prior indicator (usually the credibility of the seller) that might identify them and indicate prior asymmetric information.

Akerlof's category of automobile lemons is more demanding. These are cars fresh out of the showroom. They ought to be ‘as good as new’, and any serious defects (‘lemons’) cannot be attributed to wear and tear. Such cars are born defective. What is needed, and offered here, is a test of depreciation of almost-new cars.

One of Akerlof's predictions is that such cars could suffer from exceptional depreciation because of this undetectable risk. But such a risk was almost entirely hypothetical, since such cars were covered by factory warranties. The hapless buyer could take the car back to the vendor, and had little incentive to pass the problem on to another buyer. The factory (or a dealer, for older cars) was much better placed to absorb this risk than any private individual, and was able to take measures to prevent it.

If lemons were uncommon (not considered by Akerlof), then they would have little impact on prices. If the risk were a large one, it might be captured in the size of exceptional factory markups, reflecting the cost of making such defects good. Alternatively, if one thinks that ‘once a lemon always a lemon’, then the risk ought to continue to impinge on the price of cars before any serious wear and tear takes place, e.g. during their first year. In particular, this would be the case before 1960 in the USA, when factory warranties only ran for three months. This is worth testing for, especially since several studies (described in Section 2) appear to have found evidence for exceptional price declines during the first year. Another source of exceptional depreciation was styling and mechanical innovation, which could make some older models relatively less attractive.

The market is investigated with samples of American and British car prices in the 1950s and the 1960s. This period mostly antedates Akerlof's article, and matches the dynamics of car pricing of his time more closely than would a study of current prices.³ It stops at the oil crisis of 1973, a turning point for car marketing and design. Section 2 below contains the argument, Section 3 describes the data, Section 4 considers the evidence from depreciation, Section 5 the evidence from markups, Section 6 compares the effect of styling on American and European cars, and Section 7 concludes.

	2 Argument

TOP Abstract 1 Introduction 2 Argument 3 Data 4 Depreciation 5 Dealer markups 6 The influence of... 7 Conclusion Supplementary material Acknowledgements References

 
An initial exceptionally steep decline in used-car prices is commonly treated as an established fact, and taken as such by Akerlof. He set himself to rebut the proposition that such decline was due to the exhaustion of a novelty premium attributed to ‘the pure joy of owning a "new" car’ (Akerlof, 1970, p.489). His article did not extinguish this argument. Frank C. Wykoff argued that new cars were a ‘superior good’ with exceptional ‘freshness’, novelty and reliability (Wykoff, 1973, p.388). Used cars were not credible substitutes for new ones (Smith, 1975, p.4–5; Porter and Sattler, 1999). In 1973, Wykoff provided empirical evidence: ‘new cars depreciated at almost twice the rate of used cars ... after the first year cars appeared to depreciate at a constant rate.’ In 1989 he was more precise: ‘First-year depreciation ranges from 35% to 40%. Second-year depreciation is about 20% and depreciation is approximately constant thereafter.’ (Wykoff, 1973, p.379; Wykoff, 1989, p.260). To Howard Rachlin and Andres Ranieri, the initial steep decline suggested hyperbolic discounting (Rachlin and Ranieri, 1992, p.96). Indeed, such a kinked curve is analogous to the ‘quasi-hyperbolic’ discount curve (steep in the first period, flatter in subsequent ones) which has been adopted in studies of time-inconsistency (Laibson et al., 2003, p.520–2).

The reason for the perception of a steep initial decline is mundane. Car dealers offered two sets of prices: one was the price at which they sold (the ‘retail’ price), the other at which they bought (the ‘wholesale’ one).⁴ There were two separate depreciation vectors, one for retail prices, and one for wholesale prices. The steep decline in car prices at the outset was simply the shifting of the car's price downwards from the retail to the wholesale schedule, combined with the normal depreciation over time. This is shown schematically in Fig. 1.

View larger version (14K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 1. Schematic depreciation schedules of passenger cars

 
In Fig. 1, the difference between the retail and wholesale price vectors at any time makes up the dealer markup. This explanation of initial price declines does not invalidate Akerlof's insight of adverse selection, but rather it demonstrates the cost of one of his proposed solutions, the dealer warranty. The dealer markup covered both the core cost of selling cars, and the cost of a warranty for older cars (not covered by the factory). Selling entailed a range of costly services: finance and storage of retail stock and spare parts, transport, advertising, display, negotiation, trade-in, finance and insurance, and service facilities (Harless and Hoffer, 2002, p.272–4). In the 1960s, dealers reported that about 73% of their selling costs were fixed, i.e. covered labour and premises.⁵ The warranty cost was not necessarily incurred ex-post, but was more commonly the pre-sale cost of reconditioning (Northwood Institute, 1967, p.153–6; Lacko, 1986).

Once the car was sold, the value to the buyer of these dealer services was largely exhausted, and could not be passed on to any new buyer. Its resale value (the ‘wholesale’ price, at which a dealer would repurchase) only represented the intrinsic services embodied in the car itself, such as conveying passengers, signalling status, and sensual pleasure.

For new car buyers, the risk of ‘lemons’ was covered by the factory warranty. This gave the factory a strong incentive to avoid lemons. For used cars, the cost of explicit or implicit warranties was largely borne by the dealer, either the risk of purchasing a lemon himself, or that of meeting customer claims. When a used car was offered in trade-in, it was either disposed of immediately at wholesale prices, or it was tested, and then reconditioned before being sold at retail. Dealers had sufficient knowledge of the market to understand the risks and rectify them (Northwood Institute, 1967, ch.6; Genesove, 1993). In one survey, warranties added about 10% to the cost of a used car, but ‘average quality did not differ between warranted and unwarranted cars.’ Warranties could be seen as insurance for worried customers, but not as a signal of quality (Lacko, 1986, p.64). Some risk-averse customers incurred this expense, which adds some variance to market data.

Private sellers and buyers could sometimes negotiate a better price with each other than with dealers, by forgoing the warranty element. That is acknowledged in British car price manuals, which have often included columns for private sales, with prices intermediate between dealers’ retail and wholesale prices (e.g. Parker Mead Ltd, 1996). In Lacko's 1979–80 sample, there was no difference in price, and no quality differences, between such cars sold privately and those sold by dealers (for cars less than eight years old). Dealers were advised to monitor private-sale ad prices on a daily basis as a guide to the market (Northwood Institute, 1967, p.162). A small majority of cars were traded privately, suggesting that adverse selection was not a serious problem (Lacko, 1986).

In a competitive market the size of the markup can be taken as a measure of the cost of selling, including quality assurance. Our first argument is that: (i) initial depreciation was not exceptional. This alone would be enough to dispense with the Akerlof model. We go on to explain how the market actually worked in the face of uncertainty, and make these further hypotheses about the remainder of a car's life; (ii) the dealer markup on used cars should be larger than on new ones, since the risk was now borne by the dealer, not the manufacturer, and markups should increase with age as selling costs were inelastic; (iii) makes with a good reputation for quality might have lower depreciation and markups; (iv) in addition, but separately, frequent re-styling would be associated with higher depreciation, since it increased uncertainty about the fashion value of older cars. These hypotheses are investigated below.

	3 Data

TOP Abstract 1 Introduction 2 Argument 3 Data 4 Depreciation 5 Dealer markups 6 The influence of... 7 Conclusion Supplementary material Acknowledgements References

 
Our data consists exclusively of car prices. Observations are annual. American used car prices were derived from the monthly National Automobile Dealers Association, Official Used Car Guide, Eastern edition (NADA). This was compiled from regular transaction price reports of member dealers and from auction sales reports, and may be regarded as sample means of the larger population of dealer and auction transactions (NADA; Northwood Institute, 1967, p.156). Prices were taken from the January or February issues, and reflected the first quarter of the model year, which began in October.⁶ This sampling date maximizes the novelty effect on prices, and reduces ambiguity about the precise age of year-old cars. The British motor industry was not comparable during the 1950s, when it expanded rapidly after post-war shortages. By the 1960s cycles were similar to American ones (Foreman-Peck et al., 1995, Table 4.1, p.94). The British source, Glass's Guide, was similar to the American one, but covered the whole country. New-car prices in the USA included federal taxes, and purchase taxes in Britain. At one-third of the new price, British taxes were about three times as high as American ones. Unlike American cars, British models might be modified at any time, but starting in 1963, the vintage was indicated by means of a letter on the licence plate, which incremented upwards for new cars every January. In 1967, the new-year letter date was moved to August. Glass continued to take its representative car as being ‘first registered in the spring’, and the sampling month is April throughout.⁷ In both manuals, used cars are defined as being in good condition, but that does not exclude opportunistic behaviour, since ‘lemons’ were by definition not easily detectable (Northwood Institute, 1967, p.157; Genesove, 1993, p.646–7).

For new cars, suggested retail prices are used, as quoted in NADA and Glass. New car list prices could be discounted substantially, and actual transaction prices are not available.⁸ But USA Ford staff calculated in 1958 that customer discounts were offset by delivery costs and state taxes. Hence the list price is a reasonable proxy for actual new-car transaction prices at that time.⁹ List prices placed a ceiling on transaction prices. Using them biases new prices upwards, i.e. in favour of Akerlof's argument: any lower actual price means that there is less of an initial discount to explain. In the Ford 1958 report, the dealer markup came to 12% of the list price, 13.6% of the factory wholesale price. Another estimate of new car markups (for the 1960s) places them higher, but provides no source.¹⁰ An authoritative source reported an average gross dealer markup of 14.7% of total sales (standard dev. 0.44) in 1960–66.¹¹ In 1987, dealer new car markups before discounts could range between 6% and 14% of list price (Harless and Hoffer, 2002, p.271). Where new-car wholesale prices are used below, a 12% of list price dealer markup is used to calculate them.

The total number of depreciation observations is 4,032. Two different sampling frames are used, each repeated three times, on different samples. The first frame is a cross-section, sampling the 1957 prices of the four previous model vintages. The second frame is a panel, which follows the 1957 vintage as it aged over five subsequent years. The initial United States 1957 sample consisted of some 54 Ford and Buick models. The same makes were sampled again in 1968 (panel) and 1969 (cross-section). Concurrent British samples of 18 Ford and Morris models were also taken (1968 and 1969), as well as United States prices of the Volkswagen ‘beetle’ sedan (one model, 1954–1973) (see Table 1).¹² This provides coverage from the early 1950s up to the early 1970s. Ford and Buick sold staple mass-market models, with distinctive positions in the USA status hierarchy, Ford as a ‘low price’ car, Buick as a ‘medium priced’ one.¹³ In Britain, the ranking was reversed, with Ford making more mid-market models.¹⁴ In the cross-sections, each vintage is likely to differ mechanically and stylistically. In the panels, cars are uniform throughout (except for wear and tear), but are observed at different ages.

View this table: [in this window] [in a new window]   Table 1 Sample models, new car prices in $US (constant 1957 prices)

 

	4 Depreciation

TOP Abstract 1 Introduction 2 Argument 3 Data 4 Depreciation 5 Dealer markups 6 The influence of... 7 Conclusion Supplementary material Acknowledgements References

 
If adverse selection was peculiar to new cars, then cars should have depreciated exceptionally during their first year. If not, they would have depreciated about the same as older cars, or even less. Our task here is simply to show that contrary to claims in the literature (see Section 2 above), and to one possible interpretation of Akerlof, cars did not depreciate exceptionally during the first year. We provide what are essentially descriptive statistics. They do not attempt to explain precise depreciation changes from year to year (for which it would be appropriate to control for vintage, model, and time fixed effects), but merely to establish that depreciation levels were not exceptional in the first year. Contrary to the ‘novelty value’ hypothesis quoted in Section 2 above, initial depreciation was in fact usually smaller than depreciation in subsequent years.

The main results are presented in Fig. 2. Three measures of depreciation are estimated, retail to retail (RR), retail to wholesale (RW), and wholesale to wholesale (WW). Of the three measures, the retail-to-wholesale depreciation rate (RW) comes closest to Akerlof's example of a car sold new by a retailer, and offered for sale soon afterwards privately. The gap between sale and resale in our case is a whole year from new – that is as close as the data allows us to get: prices for younger used cars are not quoted in the source, which suggests a thin market, though not necessarily for the reasons suggested by Akerlof. Typically, new cars will have lost 30 to 40% of their value in transit from the dealer to the purchaser during their first year. Such a large absolute loss on an almost-new car would in itself deter sellers quite independently of any risk of ‘lemons’, even if they had a legitimate reason to sell so shortly after completing a major purchase. In any case, as we shall see, this loss was no greater in relative terms than the one incurred by a private individual when re-selling an older car shortly after buying it.

View larger version (28K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 2. Annual automobile mean depreciation rates by age, USA and UK. Sources: Appendix, Tables A1, A2

 
The cars that dealers sold (retail) were typically offered with some form of warranty, while those that dealers purchased (wholesale) were not. Most used-car dealers accepted some responsibility for the higher-value cars they sold, though the warranty terms varied from dealer to dealer (Northwood Institute, 1967, p.169–70). Retail-to-retail (RR) was depreciation of cars defined as being in good condition, without adverse selection. It simply measures the year-to-year depreciation of the retail offer price. Wholesale-to-wholesale price depreciation (WW – reported only in the Appendix tables) affected a population of cars that included potential lemons, and therefore embodied the risk of adverse selection undetectable by a dealer. Wholesale-to-wholesale depreciation was almost universally higher (though not always significantly) than retail-to-retail (Appendix Tables A1-A2).¹⁵ Wholesale price depreciation simply captured the larger quality variance among used cars in private ownership, before testing and repair.

Figure 2 shows that older cars depreciated no less than newer ones, and usually depreciated more.¹⁶ Without controlling for model, cohort and period attributes (for which adequate data were not available in this source), it is impossible to explain precisely the changes in depreciation rates. But our argument is about levels, not changes, and appears to be robust: whatever sampling frame is used, there is no evidence that initial depreciation rates were uniquely exceptional, and it would require an implausible set of confounding effects to reject this finding. In some years depreciation levels declined somewhat in the second year, but not enough to sustain the initial ‘novelty value premium’ argument. This is echoed in Wykoff's large sample of 1980s business-lease cars: he found an apparent ‘lemon’ effect only in one small section of his overall sample, the Oldsmobile Delta models. Their price at 18 months was higher than at six months. But the effect was weak, and suggests some issue particular to that model (Wykoff, 1989, pp.267, 289, n.11).¹⁷

In Britain, depreciation rates were higher during the first year than in the United States. In the second year they dropped sharply and only regained initial levels at age six (Appendix Tables A1, A2, (c)). This would seem to support the Akerlof model. But these UK year-two values appear to be exceptional outliers, associated with the change in the new model licence plate date letter from January to August in 1967. The dip of 0.11–0.12 points from age one to age two is probably untypical. In the four preceding years (1965–68) the cross-section second year dip averaged only 0.027. These higher initial depreciation rates in Britain remain out of line with our argument, and may indeed display a local ‘lemon’ behaviour. If that is indeed the case, then the most likely explanation is that new list prices were discounted heavily in new-car purchase transactions. This is plausible given the large share of fleet sales in Britain. Genuine adverse selection could also arise if British new-car quality was exceptionally poor, for which there is indeed ample evidence (Whisler, 1999, p.327–49, 358–9). By year two, a ‘lemon’ would be fixed, or could not be so easily disguised, and depreciation levels could then fall substantially to their long-run equilibrium. But the ‘lemon’ interpretation is doubtful even in these cases: in both these instances, higher initial depreciation is also observed in retail-to-retail depreciation, which should largely be immune to adverse selection. This makes it more likely that what we observe is a cohort or period effect, and weakens the case for regarding it as evidence for ‘lemons’. British depreciation rates were usually about six percentage points less than American ones during the 1960s. Due to higher taxation, British cars were made at even lower cost than price differences suggest, and less efficiently as well. Why they held their values better is considered later.

The six sampling frames confirm that depreciation rates usually increased in time, with lower depreciation rates in Britain (except for year one). But there were exceptions: apparent first-year ‘lemon’ effects were also evident in the USA Ford cross-section of the late 1960s. In general, therefore, higher initial depreciation rates were only seen in a minority of cases. At the most, a typical ‘lemon’ pattern (exceptional depreciation in year one) would sometimes emerge under local conditions, but was not the common pattern.

	5 Dealer markups

TOP Abstract 1 Introduction 2 Argument 3 Data 4 Depreciation 5 Dealer markups 6 The influence of... 7 Conclusion Supplementary material Acknowledgements References

 
An alternative measure of Akerlof's initial depreciation is the size of dealer markups. Dealer markups are defined here as the difference between the retail price and the wholesale price of a particular model in a particular year, expressed as a percentage of the retail price. This is different from depreciation, which compared current car prices to those in the previous year. Apart from almost-new cars fresh out of the showroom (a special case for which no used-car-wholesale-price is available), this is a better measure of Akerlof depreciation – it captures the instantaneous loss of value that occurs when a car changes hands from dealer to buyer. Table 2 shows that markups increased with age as a percentage of selling price. Markups also rose faster for American than for European cars

View this table: [in this window] [in a new window]   Table 2 Mean used-car markups by age (percent of retail price)

 
Our hypothesis is that the greater the uncertainty about the condition of wholesale cars, the higher the dealer markup as a percentage of price. This is an extension to older cars of Akerlof's insight, that uncertainty about quality affected selling margins. But although markups increased with age as a percentage of selling price, in absolute terms they remained roughly constant even for cars of different initial prices and quality. This suggests that absolute markups were capped by competition, and held up by fixed selling costs.

This is tested in the regressions reported in Table 3. Two controls are added: intrinsic differences in initial quality are controlled for by a variable consisting of the price when new, and there is a control for the make of car.

View this table: [in this window] [in a new window]   Table 3 Retail car markups (dependent variable: (log) markup as percent of dealer price)

 
Our model is

(1)

Where MARKUPPC = (Price_retail–Price_wholesale)/Price_retail for model i at time t. NEW is the retail price of the car model when new. This variable is intended to capture the quality ranking of the model when new and to test for the persistence of a quality ranking premium in older cars. FORD is the make indicator variable (the reference variable is Buick or Morris for the United States and Britain respectively), and AGE is an age-of-car indicator variable. It might be desirable to test the hypothesis that depreciation affects markups, but depreciation is collinear with AGE. A separate variable for the absolute dollar value of markups was also estimated, but proved to be small, collinear, and only marginally significant, and has been dropped. The log form provides a slightly better fit, as well as a direct measure of elasticity. Cross-sections are estimated by independently pooled OLS. Panels are estimated with GLS random effects regressions.¹⁸

The regressions in Table 3 indicate that markups increase with age. That is consistent with hypothesis (ii). Hypothesis (iii), that more expensive cars will have lower markups, is also confirmed. New-car prices have a remarkably stable negative elasticity in markups of about –0.52 for three out of four samples controlling for age and make, and –0.66 for the fourth. In other words, the more expensive the car when new, the lower the used-car markup. Expensive used-car models delivered a smaller percentage margin for dealers in comparison with cheaper cars (also White, 1971, p.106). This suggests that the main determinant of markups were the fixed selling costs. Any initial novelty premium could not be maintained as cars aged.

Selling costs, in dollar terms, remained approximately constant at all ages. This is captured by the variance of absolute dollar markups within each sample. The mean coefficient of variance for all four American samples, over all ages was only 0.17 (16 observations, unweighted; min. 0.15, max. 0.21).¹⁹ As we have seen, about three-quarters of new dealership costs were fixed, i.e. spent on staff and premises (n.5 above). In a competitive market without cross-subsidy, fixed costs caused percentage markups to be low on newer cars, and high on older ones. Dealers could and often did cross-subsidize, e.g. overpricing trade-ins in order to sell new cars, but the handbook car prices that we use rest on the bedrock of realisable market value.

Quality uncertainty and the fixed costs of selling, jointly placed a limit on the ability of dealers to sell older cars. Table 4 is derived from a Federal Trade Commission nationally-representative survey of car buyers in 1979–80 (Lacko, 1986; Genesove, 1993). It shows the percentage of cars of every age sold in three types of transactions: by new car dealers, by exclusively used car dealers, and by private individuals. In this sample, used cars less than one year old were traded, though in smaller numbers than older ones. Four-fifths of cars under one-year old (25) were sold by new car dealers, who alone could provide credible warranties. But one-fifth (7 cars) were sold privately. Dealers facilitated exchange, but a few private individuals also found the confidence to trade. It was the high transaction cost (some 20 to 40% of a new car's price) that must have largely discouraged trade in such cars.

View this table: [in this window] [in a new window]   Table 4 Market shares of used cars by age, USA, Dec. 1979

 
Five-year-old cars had markups in excess of 30% of shrinking retail prices (Table 2 above). As these cars depreciated, the markups required to recondition, sell, and guarantee the cars increased as a proportion of their price. Private sellers had few of these costs. Consequently, new car dealers largely withdrew, and customers assumed the risk themselves. Self-insurance was a reasonable choice since older cars (more than five years) were cheap, having lost some four-fifths of their initial values. Some business was taken over by dealers specialising in used cars (whose facilities and warranties would be more limited), but for older cars overall, the high transaction costs priced out intermediaries, and the bulk of such cars were traded privately (Table 4). It was only at this stage that the ‘lemon’ risk was finally shifted from seller to buyer. Interestingly, in some other studies, it is only in the oldest cars (more than ten and seven years old respectively) that any adverse selection effects could be detected (Bond, 1982; Lacko, 1986).

Table 4 shows that the almost-new market was not thin. There was manifestly an incentive to ignore the risk of adverse selection and to absorb high transaction costs in order to dispose of such cars, where that was necessary. There is other evidence as well: in the 1950s, 5–6% of new cars were re-sold in less than a year, and another 10–11% after one to two years (Crowell-Collier, 1955, p.18; Look, 1960, p.26). In the 1960s, most rental cars were sold after a year (White, 1971, p.169). In a 1979–80 survey, 7.5% of used cars sold by new-car dealers were less than one year old (Genesove, 1993, Table 1, p.651, col. 1). Of a very large sample of business-lease cars in the 1980s, one third were re-sold within the first year (calculated from Wykoff, 1989, Table 6.2, p.270–1). Trade was manifestly not disabled. Any problem of adverse selection appears to have been solved by the market through the use of dealer guarantees.

British cars differed from American ones: The new car price ranking (Table 3, variable (LOG)NEW, cols. 5, 6) persisted for older car markups, i.e. the coefficient was not significantly different from zero. And the age markup increment was much lower in comparison with American cars. In other words, British cars depreciated much less than American ones, and British dealers earned lower markups.

	6 The influence of styling

TOP Abstract 1 Introduction 2 Argument 3 Data 4 Depreciation 5 Dealer markups 6 The influence of... 7 Conclusion Supplementary material Acknowledgements References

 
Car price manuals do not provide usable evidence on the hedonic attributes of particular cars, either mechanical or styling ones. Some evidence of a styling effect can nevertheless be detected in a comparison of American, British, and USA Volkswagen depreciation rates. American and European manufacturers followed different styling strategies. During the 1950s, the dominant American manufacturers pursued a frenzied styling race by means of annual face-lifts and body make-ups (Fisher et al., 1962; Offer, 1998). This slowed down somewhat during the 1960s (Pashigian et al., 1995, p.291–2). In Britain, in contrast, the styling cycle was much slower. Ford renewed their models about every four years, while Morris kept models in production for at least a decade, and sometimes much longer.

In Britain, the annual novelty signal was a letter on the licence plate. It was uniform and costless. In the United States, novelty was signalled expensively by means of annual styling facelifts. Depreciation rates (Fig. 2 above) highlight the much more dynamic and competitive styling/innovation regime in the United States, compared with Britain. Apart from year one, British depreciation rates were about one-fifth lower than American ones, despite poor British mechanical quality. Model attrition was much faster in the United States. In the cross-sections, earlier models US were discontinued rapidly, while British ones persisted (Table A1, cross-sections).

Volkswagen USA made a virtue of a single, reliable, economical, and fixed design. Customers were credibly shielded from fashion obsolescence, a point stressed in company advertising.²⁰ Volkswagen retail-retail depreciation was exceptionally low.

Indigenous British manufacturers were also slow to change designs. The vintage letter on the licence plate provided a novelty signal instead. Table 5 shows that British cars and Volkswagen in the United States depreciated much less than comparable American cars. After five years, a British car (or Volkswagen in the USA) was worth almost 50% more than an American one of the same age, as a proportion of the original price.

View this table: [in this window] [in a new window]   Table 5 Wholesale prices as percentage of new prices, American and European makes

 
It is unlikely that British cars were more durable—they failed in the American market while Volkswagen succeeded. They were built at much lower cost. They had poor quality reputations (Whisler, 1999, p.327–49, 358–9). The low variance in Glass's Guide might suggest that the prices reported might have been derived from a formula, rather than reported empirically, but the strong Ford cohort effect in age three (Table A2) suggests otherwise, and so does the variance in 2^nd year UK depreciation over several years. Glass had an incentive to get it right.

For both the UK models and for Volkswagen in the USA, there was less uncertainty about fashion durability. In the United States, with its styling race, makers could not credibly commit not to innovate (Coase, 1972; Purohit, 1992). Carmakers in Britain (and Volkswagen USA) could do so. That was reflected in lower depreciation, and lower markups. Although styling strategy was unlikely to be the only source of depreciation level difference between the two markets, the evidence remains suggestive. The ‘folk wisdom’ which explained initial price declines by the exhaustion of novelty, appears to have contained a kernel of truth.

	7 Conclusion

TOP Abstract 1 Introduction 2 Argument 3 Data 4 Depreciation 5 Dealer markups 6 The influence of... 7 Conclusion Supplementary material Acknowledgements References

 
Akerlof's model of almost-new car prices is empirically indeterminate. It is consistent with three different outcomes: (i) no trading, (ii) trading with depressed prices, or (iii) trading with dealer warranties. He did not consider that new-car ‘lemons’ might be sufficiently rare so as not to impinge on car trading. Factory-fresh used cars were a special case, hardly ever sold, but not necessarily for adverse-selection reasons. For cars during their first year, dealer warranties appear to have enabled trading. The stylized fact of exceptional initial depreciation, accepted by many previous writers, is explained by their overlooking the shift from a retail to a lower wholesale depreciation curve when a vehicle was sold. The actual pattern found is that (i) cars in their first year were traded, normally by dealers; (ii) initial depreciation was no higher, and was usually lower, than depreciation of older cars. This is confirmed by analysis of dealer markups. Dealer services were genuine transaction costs, reflecting the real cost of doing business in a competitive market. As cars aged, fixed costs increased as a percentage of value, until full-service dealers could no longer afford to trade. Earlier studies have found that adverse selection was only weakly present (if at all) in older cars. It is now shown that adverse selection left no enduring mark on the price of cars at the end of their first year either. In general, the evidence for the widespread effect of ‘lemons’ is weak. It suggests that ‘lemons’ were not a pervasive hazard, although they might appear in particular local circumstances. As an empirical matter, the used car market may not have been the best example to use in demonstrating the effect of ‘lemons’. Popular intuitions can be wrong.

	Supplementary material

TOP Abstract 1 Introduction 2 Argument 3 Data 4 Depreciation 5 Dealer markups 6 The influence of... 7 Conclusion Supplementary material Acknowledgements References

 
Supplementary material (the Appendix) is available online at the OUP website.

	Acknowledgements

TOP Abstract 1 Introduction 2 Argument 3 Data 4 Depreciation 5 Dealer markups 6 The influence of... 7 Conclusion Supplementary material Acknowledgements References

 
Thanks to Annie Chan, Harriet Jackson, Maïa-Laura Ibsen, and Siobhan McAndrew for painstaking data collection; to Gavin Cameron, Paul David, Charles Feinstein, Siobhan McAndrew, Mara Meacci, Raphael Schapiro, Peter Temin, and to seminar participants at Oxford and Rome, for helpful advice; to John Muellbauer for acute comments and for suggesting the title, and to this journal's referees for guidance and goodwill.

	Notes

 
¹ Omitted in the British edition.

² In contrast, theoretical studies argue that adverse selection ought to exist (Kim, 1985; Hendel and Lizzeri, 1999).

³ Pashigian et al. (1995) have shown that these dynamics are not stable over time.

⁴ New wholesale prices were not reported. (National Automobile Dealers’ Association, 1957–73; Glass's Guide Services Limited, 1968–73).

⁵ Calculated from ‘Auto Dealers Sales Expense and Operating Profit Before Federal Income Tax’, (Ward's Reports, 1960–1966).

⁶ Buick 1969 models, from December 1968. Including heater and radio. This source is difficult to obtain. A single run, starting in 1957, was located at the New York Public Library.

⁷ Following practice in the motor trade (Glass’s Guide, Nov. 1969, p.2–4, cited from p.4, col.5).

⁸ Ward's Reports (1960, p.123), and Pashigian (1961, p.37) report excess capacity after 1954.

⁹ On the basic 1958 Ford car. List price $1977 from NADA, (Feb. 1958); transaction cost ($1994) calculated from Ford Product Planning Office, Economy Car Report, Nov. 13 1957, Ford Industrial Archives AR-94-200777-5, fo.33; Product Planning Committee, 13 Nov. 1957, Ford Industrial Archives, AR-94-200777-6. The nominal list price (full dealer markup; state taxes and freight excluded) was $1935. The original figures are adjusted to the NADA bare car price basis by removing optional accessories, and adjusting taxes and the reported dealer markup proportionally (Ford Motor Company (1957a, b).

¹⁰ 18% of retail (17–25% of wholesale), increasing in price, and including more expensive cars. White (1971, p.106).

¹¹ Ward's Reports, ‘Auto Dealers Sales Expense and Operating Profit Before Federal Income Tax’, (1961–7).

¹² Sampling was constrained by lack of data before 1957: the cross-section method gave access to earlier years. It was limited to four years (six in the UK) due to model attrition. In the panels, vehicles lost about 80% of their value after five years. 1968 (panels) and 1969 (cross-sections) were chosen in order to avoid 1967, when model year date indicators changed in Britain, and to fit in five panel years before 1973.

¹³ Offer (1998) discusses model hierarchies.

¹⁴ Morris models were produced by the British Motor Corporation (British Leyland from 1968), which sold an identical ‘Austin’ line as well.

¹⁵ Appendix Tables A1 and A2 are posted on the OUP website as Supplementary material.

¹⁶ Also observed by Purohit (1992, p.164) for 1975–85 cohorts in the USA.

¹⁷ Lower depreciation increases in the 1960s may be consistent with quality improvements consequent on an extension of factory warranties from three months to one year in 1960.

¹⁸ Cross-section time-series panel estimates corrected for autocorrelation and with robust-standard-errors have also been tested, and make only very small and insignificant differences to coefficients and errors. Weighting by sales is not practical, since the data (for an American sub-region and for particular sub-periods) would be impossible to obtain. The cars are all mass-production lower or middle-market models, and it would require a dominant model with strong outlying characteristics to reverse the results. Initial characteristics are controlled for by means of the variable NEW.

¹⁹ Two makes x (four cross-sections + four panels).

²⁰ Volkswagen USA ads: http://www.ciadvertising.org/studies/student/99_spring/interactive/joohwan/bernbach/images/vwad15.gif. Accessed 25 March 2003. Copy in possession of the author.

	References

TOP Abstract 1 Introduction 2 Argument 3 Data 4 Depreciation 5 Dealer markups 6 The influence of... 7 Conclusion Supplementary material Acknowledgements References

 

Akerlof G-A. The market for ‘lemons’: quality uncertainty and the market mechanism. Quarterly Journal of Economics (1970) 84:488–500.[CrossRef][Web of Science]

Bond EW. A direct test of the ‘lemons’ model: the market for used pickup trucks. American Economic Review (1982) 72:836–40.[Web of Science]

Bond EW. Test of the lemons model: reply. American Economic Review (1984) 74:801–04.[Web of Science]

Coase R. Durability and monopoly. Journal of Law and Economics (1972) 15:143–9.[CrossRef][Web of Science]

Crowell-Collier Publishing Co. Crowell-Collier Automotive Survey (1955) 19.

Emons W, Sheldon G. (2002) The market for used cars: a new test for the lemons model, Discussion Paper No. 3360: Centre for Economic Policy Research, London.

Fisher FM, Griliches Z, Kaysen C. The Costs of automobile model changes since 1949. The Journal of Political Economy (1962) 52:433–51.

Ford Motor Company. Product Planning Committee, Ford Industrial Archives, AR-94–200777–6, Detroit. (1957a).

Ford Motor Company. Product Planning Office, Economy Car Report, Ford Industrial Archives, AR-94–200777–5, Detroit. (1957b).

Foreman-Peck J, Bowden S, McKinlay A. The British Motor Industry (1995) Manchester: Manchester University Press.

Genesove D. Adverse selection in the wholesale used car market. The Journal of Political Economy (1993) 101:644–65.[CrossRef]

Glass's; Guide Service Limited. Glass's Guide to Used Car Values (1968–73) Weybridge: Glass's Guide Service Limited.

Harford T. The Undercover Economist: Exposing Why the Rich Are Rich, the Poor Are Poor—and Why You Can Never Buy a Decent Used Car! (2005) New York: Oxford University Press.

Harless DW, Hoffer GE. Do women pay more for new vehicles? Evidence from transaction price data. American Economic Review (2002) 92:270–79.

Hendel I, Lizzeri A. Adverse selection in durable goods markets. American Economic Review (1999) 89:1097–115.[Web of Science]

Kim JC. The market for ‘lemons’ reconsidered: a model of the used car market with asymmetric information. American Economic Review (1985) 75:836–43.[Web of Science]

Lacko J. Product Quality and Information in the Used Car Market (1986) Bureau of Economics, Washington, DC: Federal Trade Commission.

Laibson DI, Rebetto A, Tobacman J, Weinberg S. The hyperbolic consumption model: calibration, simulation and empirical evaluation. In: Time and Decision: Economic and Psychological Perspectives on Intertemporal Choice—Leowenstein G, Read D, Baumeister RF, eds. (2003) New York: Russell Sage Foundation. 517–43.

Look. In: National Automobile and Tire Survey (1960) 24. New York.

National Automobile Dealers' Association. Official Used Car Guide, Eastern Edition (1957–1973) Washington, DC: National Automobile Dealers’ Association.

Northwood Institute. Merchandising Cars and Trucks (1967) MI: Midland.

Offer A. The American Automobile Frenzy of the 1950s. In: From Family Firms to Corporate Capitalism: Essays in Business and Industrial History in Honour of Peter Mathias—Bruland K, O’Brien PK, eds. (1998) Oxford: Clarendon Press.

Parker Mead Ltd. Parker's Car Price Guide (1996) December, London.

Pashigian BP. The Distribution of Automobiles, an Economic Analysis of the Franchise System (1961) Englewood Cliffs, NJ: Prentice-Hall.

Pashigian BP, Bowen B, Gould E. Fashion, styling, and the within-season decline in automobile prices. Journal of Law and Economics (1995) 38:281–309.[CrossRef][Web of Science]

Porter RH, Sattler P. (1999) Patterns of trade in the market for used durables: theory and evidence, Discussion Paper No. 7149: NBER, Cambridge MA.

Purohit D. Exploring the relationship between the markets for new and used durable goods: the case of automobiles. Marketing Science (1992) 11:154–67.[Abstract/Free Full Text]

Rachlin H, Raineri A. Irrationality, impulsiveness, and selfishness as discount reversal effects. In: Time and Decision: Economic and Psychological Perspectives on Intertemporal Choice—Leowenstein G, Read D, Baumeister RF, eds. (1992) New York: Russell Sage Foundation.

Smith RP. Consumer Demand for Cars in the USA (1975) Cambridge: Cambridge University Press.

Sugden R. Credible worlds: the status of theoretical models in economics. Journal of Economic Methodology (2000) 7:1–31.[CrossRef]

United States Council of Economic Advisers. Economic Report of the President (1991) Washington, DC: United States Council of Economic Advisers.

Ward's; Reports Inc. Ward's Automotive Yearbook (1938) Detroit: Ward's Reports Inc.

Whisler TR. The British Motor Industry, 1945–1994: A Case Study in Industrial Decline (1999) Oxford: Oxford University Press.

White LJ. The Automobile Industry since 1945 (1971) Cambridge, MA: Harvard University Press.

Wykoff FC. A user cost approach to new automobile purchases. Review of Economic Studies (1973) 40:377–90.[Medline]

Wykoff FC. Economic depreciation and the user cost of business-leased automobiles. In: Technology and Capital Formation—Jorgenson DW, Landau R, eds. (1989) Cambridge, MA: MIT Press.

CiteULike    Connotea    Del.icio.us    What's this?

This Article Abstract FREE Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Alert me to new issues of the journal Add to My Personal Archive Download to citation manager Request Permissions Google Scholar Articles by Offer, A. Search for Related Content Social Bookmarking          What's this?

Online ISSN 1464-3812 - Print ISSN 0030-7653

Copyright © 2009 Oxford University Press

Oxford Journals Oxford University Press

• Site Map
• Privacy Policy
• Frequently Asked Questions

Other Oxford University Press sites: Oxford University Press Oxford Journals China Oxford Journals Japan American National Biography Booksellers' Information Service Children's Fiction and Poetry Children's Reference Corporate & Special Sales Dictionaries Dictionary of National Biography Digital Reference English Language Teaching Higher Education Textbooks Humanities International Education Unit Law Medicine Music Online Products Oxford English Dictionary Reference Rights and Permissions Science School Books Social Sciences Very Short Introductions World's Classics