Test of the Lemons Model: Comment By Micuaet D. PRatr AND GEORGE E. HOFFER* In a recent article in this Review, Eric Bond (1982) provided an empirical test of the “lemons” model. Using the 1977 Truck Inventory and Use (TIU) Survey, Bond de- termined that there was not a significant difference between the maintenance records of pickup trucks which were purchased new and those purchased as used vehicles. He suggests several explanations for his findings. ‘These include the counteracting institutions discussed by George Akerlof (1970), as well as the ability of consumers to obtain enough information to eliminate the buyer-seller in- formational asymmetry. While Bond’s con- clusions are interesting, we propose to pro- vide a finer test of the lemons model in order to determine whether these conclusions accu- rately reflect conditions in the market for used pickup trucks. Our model introduces two refinements to Bond’s analysis We believe that Bond was unable to verify the lemons model because of the lack of expenditure data in the TIU Survey. As he notes, the quality of a used vehicle depends “not only on the probability of maintenance but on the costliness of repairs” (p. 837). However, the relative cost of repairs cannot be captured to any significant degree in the maintenance categories of the T/U Survey. ‘As shown in Table 1, the four defined main- tenance categories range in average expendi- ture from $100 to $1,200. Therefore, Bond's tests of differences in frequencies of repairs between vehicles acquired new and those acquired used are not as fine a test as neces- sary to verify the lemons model. Not only does this test fail to capture the relative cost difference between repairs, it cannot account for vehicles that have more than one major repair during the year—perhaps the true Jemon, *Virginia Commonwealth University, Richmond, VA 23284, We thank Eric Bond, Robert Reilly, and an anonymous referee for their helpful comments. "We thank the referee for this interpretation. 798 ‘TaBLe 1—RELATIVE REPORTED MAINTENANCE Costs FOR PICKUP TRUCKS, 1977 Brakes 5 100 Rear End—Differential 200 Transmission 400 Engine 1200 Source: Automotive Service Council of Virginia (1983). Moreover, the Bond model does not cap- ture the complete character of a lemons market, As noted by Akerlof, it is more likely that the owner of a lemon will sell his vehicle than the owner of a “creampuff.” For if a lemons market is presumed to exist, the owner of the creampuf vehicle will not be rewarded for the superior condition of his vehicle. Therefore a comparison of trucks acquired new versus trucks acquired used would identify only those vehicles for which the lemon characteristic is permanent.> One should compare trucks purchased used dur- ing a specific time period with those held during that time period, whether they had been acquired as new or used vehicles, in order to capture both the “ permanent lemon” as well as the “ transitory lemon.” The latter vehicle is the recently purchased used vehicle which is repaired by the new owner. In the model developed below, we test the lemons hypothesis using the 77U Survey data taking into account the two elements of a lemons market that should be included for a finer test of this hypothesis. I. The Model To reflect the essential time clement and market process described by a lemons model, the TIU Survey data were segmented into two transaction groups.’ In the first group 2We thank Bond for this clarification. Bond's analysis involved pickup trucks aged 5 years or less. We have included all pickup trucks in. our analysisVOL. 74 NO.4 we included only pickup trucks that had been purchased used within one year of a state’s survey date. These are the vehicles which would be more likely to include Iemons. Our second group included pickup trucks which had been acquired either as new vehicles or used vehicles in some previ- ous period and kept by that present owner. If there is a lemons market, this group should contain relatively more creampuffs. Excluded from both groups were vehicles that were reported to be leased, to be not in use, or to have been sold during the survey period. Although Bond’s test required that he con- trol for vehicle age and total vehicle mileage, our test does not dictate this type of control. This is important, for to exclude some vehicles from the analysis because of their age and/or mileage characteristics precludes a full modeling of a lemons market.‘ At any given time, the market consists of a spectrum of trucks having various age and mileage characteristics. These characteristics provide information to a prospective buyer of a used vehicle but they do not effect the buyer-seller informational asymmetry. To capture the relative cost of the reported maintenance categories we used the cost data presented in Table 1.° For example, as shown, in 1977 the average expenditure on transmis- sion maintenance per year was four times that of the average expenditure on brake ‘maintenance. If the market for used vehicles is a lemons market, we would expect that the average “Bond! tests requite that he segment the market for used vehicles using age and mileage as. technical device. In the marke, there sa segmentation of vehicles into sets that are considered by buyers and sellers to be nearly perfect substitutes, This need not correspond to a technical segmentation by age and lifetime’ mileage Hence, although all trucks in Bond's sample are con- sidered in his tess, this echnical segmentation may not allow a proper test of the informational asymmetry tpetween buyers and Sellers within the same market segment “Because of obvious ambiguity, the category used to report “other maintenance” was excluded from the anal yas, We would expect that this category would include maintenance types whose expenditures would be loner than the average brake repai. Inclusion of this mainte nance category does not afet the esl PRATT AND HOFFER: LEMONS MODEL TEST 799 ‘Taste 2—Tesrs oF THE EQUAL 'Y OF MEAN "MAINTENANCE EXPENDITURES" ‘Standard Standard N Mean Deviation Error Recently Transacted 2395 $218 542M Not Recently Transacted 10.799 251 488 005 “Differences in the means are significant at the OL level yearly maintenance incurred for the four ‘major repair categories reported in the T/U Survey to be significantly different between the two vehicle groups. A priori, if the lemons hypothesis is true, we would expect that the average expenditure on vehicles purchased used within our observation period to be greater than the average maintenance expenditures on vehicles in the kept group. Under the null hypothesis that the market for used vehicles is not a lemons market, there should be no difference in the mean repair expenditures between our transaction groups. Thus a test of the equality of the mean repair expenditures between these groups is performed to determine the accep- {ance of the null hypothesis. The difference in the means reported in Table 2 is significant at the 1 percent level.” Therefore we reject the null hypothesis, and given the ranking of the mean repair expenditures, we conclude that the market for used pickup trucks is a lemons market. Therefore arguments for the presence of Akerlof-type counteracting institutions, as well as explanations for a smaller difference in buyer-seller information, although possi- There is not a significant difference between the average age of average lifetime mileage of these two transaction groups, Therefore, out test does not merely reect diferences in repair records that one might ex pect to exist de to significant differences in the average fge or average lifetime mileage of vehicles in the (WO transaction groups TA. semsitvity analysis shows that the results were invariant with respect to the actual dollar amounts reported in Table 1. The analysis is dependent on the relative ranking of the reported expenditure categories.300 THE AMERICAN ECONOMIC REVIEW bly correct for some used markets, may not be significant factors in this market." "4 ditect test of the impact of counteracting insti tions om the market for used vehicles is provided in our ear paper (1983) REFERENCES Akerlof, George A., “The Market for “Lemons’: Quality Uncertainty and the Market Mechanism,” Quarterly Journal of Econont- ies, August 1970, 84, 488-500. SEPTEMBER 1984 Bond, Erie W.,“A Direct Test of the ‘Lemons’ Model: The Market for Used Pickup Trucks,” American Economic Review, Sep- tember 1982, 72, 836-40. Pratt, Michael D. and Hoffer, George E., “The Efficacy of Used Vehicle Disclosure Laws,” Working Paper No. 83-01, Department of Economics, Virginia Commonwealth Uni- versity, 1983. USS. Department of Commerce, Bureau of the Census, Truck Inventory and Use Survey, 1977, Washington: USGPO, 1981 Automotive Service Council of Virginia, unpub- lished data, Richmond, Virginia, 1983.