Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
8 Technology and Competition in the U.S. Automobile Market The crisis in the U.S. auto industry reflects in part the vigorous competitive challenge of the Japanese, in part a rapid shift in market preferences. We have argued that a recapture of com- petitive cost and quality performance will require fundamental changes In the way the manufacturing process in the U.S. auto industry is managed. At the same time the industry is faced with the problem of developing new products to meet changing market demands. "Small" and "fuel efficient" are the words most often used to describe desired characteristics, and, however superficial those descriptions may be, it is becoming increasingly clear that competitive vehicle design requires product technology different from that found in the standard American sedan of the post-World War II period. Throughout that era, competition in any given segment of the U.S. automobile market occurred largely on the basis of economies of scale, styling, and sales and service net- works. As befits a maturing industry, innovation became increas- ingly incremental in nature and, in marketing terms, virtually invisible. Has that situation now changed? Clearly, some major changes in product technology have occurred in the last few years, most notably the introduction of front-wheel drive, the trans-axle, and the increased use of the diesel engine. Yet it is not clear whether such developments reflect the beginning of new technological ferment or the end of a technological transition. Indeed, some suggest that the small, front-wheel-drive car, with its transverse mounted, four-cylinder engine, already constitutes a new dominant design. If so, future changes in automobile technology are likely to be incremental to the new design, and competition will occur much as before on the basis of styling, scale economies, and dealer networks. This reading of events places the industry farther along its path toward maturity. If, however, innovation in technology is again becoming of vital competitive significance, not only are we likely to see con- . . .. . 122
123 tinuing functional innovations but the overall pace of innovation is likely to increase as well. These developments, in turn, may have far-reaching implications for the structure of the industry, for individual strategic decisions, and for the pattern of international trade. The question of technology's role in competi tion is thus central to interpreting current events and is an important aspect of the larger question of the industry's future. In marketing terms the notion of "competitive significance" has both a quantity and a price dimension. A given technology (e.g., front-wheel drive) becomes significant in a competitive sense either if consumers are willing to pay a premium for cars- embodying the technology or if the market share of such models increases. Accordingly, our analysis in this chapter will focus on both sales and price effects of a few major technological char- acteristics in the context of a fairly simple model of consumer demand (at the compact and subcompact end of the market). Our hypothesis is that the oil shock of 1979 altered demand patterns, thus increasing both the visibility of technology and its competi- t~ve significance. If true, we should observe very different market valuations of those characteristics before and after 1979. A FRAMEWORK FOR ANALYSIS The framework we use to test the competitive significance of technology is designed to identify the market's valuation (sales and price) of a given characteristic. There are few difficulties with sales effects. In fact the only major issue is the problem of capacity constraints. In a given year the sales of a particular model may be more a reflection of the capacity of the firm to produce it than of underlying consumer demand. But since our purpose is to estimate the average effect on sales of a given characteristic, and since each characteristic tends to be found on m ore than one model, it follows that only if all models with a given characteristic are subject to capacity constraints will sales reflect those constraints and not market demand. The sales model we use is presented in Appendix C. Our approach simply is to identify the impact of a particular tech- nological feature (e.g., front-wheel drive) by statistically holding constant the effects of other characteristics. Estimating the market value attached to specific character- ~stics is a somewhat trickier process, for it relies on a set of assumptions that must be spelled out clearly. The problem here lies in the fact that there is no market for individual character- istics as such. A given model contains by definition a bundle of characteristics or attributes, and its price in the market is the price of the bundle as a whole.
124 To identify the "price" of a particular attribute, we assume that observed prices are a reflection of an implicit market in characteristics. Consumers place a value on specific attributes that have been aggregated or bundled in different ways to form different models. To infer the value of a specific characteristic, we take a range of model prices, some of which have the charac- teristic and some of which do not. By observing how the overall price changes with variation in characteristics, we can identify the value consumers place on specific attributes. The details of this approach are presented in Appendix C. A difficulty arises, however, in interpreting the prices of these attributes. What we are after is the market's or the consumer's valuation of each characteristic, but the market price of a given m odel is determined by the interaction of supply as well as demand forces. On the supolY side the Rev determinants are the . . . . , , , _. _ .. _ costs ot production, the pricing policy (e.g., relationship of one model's price to another), and the strategy of the firm. The demand side is determined by the Dreferences of c'~n~'m~rc i ~ . by their assessments of the value of a given characteristic. Though only demand considerations are relevant for our purposes, the price of a model reflects both market valuation, which interests us, and the manufacturers cc,rtc anal nri~ina mail ire which do not. r ^~- ~ r~~~~~ One solution to the problem of isolating the demand effects is to specify a structural model of consumer demand and the costs of production. Given appropriate exogenous variables one can use advanced statistical techniques to disentangle supply and demand effects.2 But a solution much less demanding of the data may be available. The lack of identifiability of the demand function may not affect inferences about changes in demand. Costs of reduction may be more stable than consumer valuation, particu- larly over short periods of time when the sales mix of characteristics shifts rapidly. If this is true, then any differences in coefficients estimated for two closely paired years will reveal the influence of changes in consumer demand. Furthermore, the availability of both list and transaction prices may provide another means of correcting for supply side effects. Existing evidence on automobile pricing suggests that list prices are determined by product-line policy and standard costs.3 While the markup over standard cost may be influenced by strategic considerations and estimates of consumer valuation, variations in list prices are likely to reflect variations in cost. Inclusion of the list price in an equation explaining transaction prices may therefore provide a control for the supply side. It is of course possible that such a procedure will "overcontrol" and thereby obscure demand effects. But estimation with and without the list price in a comparative context should provide a basis for inferences about shifts in demand. -
125 Empirical Specification The framework developed above requires data on sales, list and transaction prices, and model characteristics that capture the effect of technology and innovation. List prices and sales data are readily available for most models sold in the United States, but transaction prices for new cars are not. As a proxy for them we have chosen to use the price of the model one year later as determined in the used-car market. Use of the year-old price introduces a further complication in the analysis, since discounts based on used-car prices will reflect the effects of deterioration as well as the market's valuation of technological character- istics. However, as long as deterioration is not a function of those characteristics, it is fair to assume that the effects of general deterioration will be reflected in the overall average for the market and will not obscure relevant model differences. The characteristics we have chosen to include in the analysis are determined by our hypothesis about changes in the role of technology in the market for automobiles. To begin, we distin- guish between performance characteristics and technological attributes, although both are closely related. Thus, miles per gallon measures fuel-efficiency performance, but the size of the engine is a characteristic of the model, and both kinds of vari- ables are included in the analysis.5 In the compact and subcompact markets on which we focus, the key performance characteristics are fuel efficiency, driving range, repair frequency, and package efficiency (efficient use of space). The relevant definitions and measures are presented in Table 8.1. The key technological characteristics include engine type (gas or diesel), drive train (front or rear wheel), and age. This last variable is meant to test the notion that newness itself is valued independent of specific characteristics. We are aware that this scheme leaves out, by necessity, a number of alternatives that may be important in the market. Since the omitted factors may be reflected in the analysis if they are correlated with variables that are included, care must be taken in interpreting the result. The drive train, for example, may pick up some of the effects of differences in handling and m aneuverability. To complete the framework we have added a set of variables that control for the country of origin and the market segment. V ariables indicating whether a car is produced in Japan (we distinguish captive imports from others), Europe, or the United States are intended to pick up any otherwise unmeasured differ- ences in quality of attributes correlated with the country of origin. Finally, we have allowed the average discount to be different in the subcompact and compact model categories. _
126 TABLE 8.1 Basic Variables: Definitions Variables Symbols and Definitions Sales Si: number of vehicles of ith model sold in specific year. List price P: fist price. Transaction price P*: one-year-old used-car price. Fuel efficiency MPG: EPA miles per gallon rating for city driving. Driving range RNG: MPG (reported fuel-tank capacity). Repair frequency (two REP1: has value of 1 if Consumer Reports' survey of repair variables entered) frequency placed model in "substantially below average" category; zero otherwise. REP2: has value of 1 if Consumer Reports' survey of repair frequency placed model in "substantially above average" category; zero otherwise. Package efficiency (two IVTV: internal volume (defined below) . total volume measures) (length x width x height). VOLWT: internal volume . vehicle curb weight. Engine type DIESEL: has value of 1 if model has diesel engine; zero otherwise. Drive train F WD: has value of 1 if model has front-wheel drive; zero otherwise. Age of model AGE: years since last major model redesign. NOTE: Using the diagrams and specifications in the figure below (taken from Consumer Reports' Annual Auto Issue, April 1980), the internal volume is obtained by adding the in- ternal volume of the front and rear compartments of the car. The internal volume of the front compartment is the product of the front shoulder room and the cross-section area of the front compartment. The internal volume of the rear compartment is similarly obtained. The formulas are as follows: Front compartment volume V1 = J [(0.5) (G - 18)2 + (24) (E) + 0.18 (E)2] Rear compartment volume V2 = K [(H- 18) (G- 18) (0.71) + (F) (H)] Internal Volume = V1 + V2 The following assumptions were made in the calculations: · distance between tester's hip and knee = 18 inches · tester's leg was inclined at 45° · seats were inclined at 10° to the vertical The dimensions E and F were defined as the height above seats rather than the clearances measured by Consumer Reports' tester. Consumer Reports' E and F were modified by adding a constant of 34 inches to obtain the heights above seat levels. Data for the E adjust- ments were obtained from Automotive News Almanac. W:~
127 - -~1~1 :,r,,. . =~ 1 . ~- ! Table 8.2 presents average values of the basic variables included in our analysis of sales and discounts. A comparison of 1977 and 1979 data reveals that most characteristics are quite similar in the two years. - Because of the introduction of new models and substantial redesign efforts, the average age fell from almost four years to a little over two and one-half. Among the other characteristics, however, we find little difference. The percentage of models with front-wheel drive Increased Slightly, as did driving range. But the changes are small, and in general the average characteristics are similar in the two years. The model-age data suggest, however. that substantial . . . . . . . . . .. . , _ changes may nave taken place both in the way in which the technology was packaged and in the quality of the technology. It is important to realize that the subcompact market in 1977 included models (in comparable numbers and at comparable levels of performance) offering the characteristics we hypothesize were relevant in the 1979-1980 market. The date suggest that a com- parison of market results in these two years might be a useful test of the role of technology in competition. Had we found gross dissimilarities in attributes, the validity of the test would have been suspect. COMPARISONS OF DISCOUNTS DURING 1977 AND 1979 To compare the impact of given attributes on the transaction price (statistical results for discounts are presented in Appendix C, Tables C.3 and C.4), we have developed what we call the basic effect of each characteristic. We first calculate how much a "typical" difference in the characteristic would have raised or lowered the price.6 For VOLWT (package efficiency), for example, a typical or average difference between a given model and the average for all models was 8.7 cubic inches per pound in 1979. We estimate that such a difference would raise the
128 TABLE 8.2 Means and Standard Deviations of Basic Variables, 1977 and 1979 1977 1979 Typical Typical Difference Difference From the From the Variable Mean Mean Mean Mean Sales per (S) model (000s) 92.6 93.4 102.8 95.1 List price (P) 4484 1209 5657 963 Transaction price (P*) 3674 901 4591 832 Fuel efficiency (MPG) 21.8 5.3 21.0 4.8 Driving range (RNG) 318.0 57.7 326.5 70.8 Package efficiency VOLWT (cubic inches per pound) 49.0 7.0 49.3 8.7 IVIV (percent) 22.5 1.7 22.3 2.8 Engine type (percept diesel) 2.3 14.9 3.5 18.5 Drive tra~n(~;WD) (percent) 25.0 43.3 27.1 44.4 Age ofmodel (AGE) 3.7 3.5 2.3 2.1 Repair frequency (percent) REP 1 15.9 36.6 3.5 18.5 REP 2 22.7 41.9 15.3 36.0 Subcompact (percent) 65.9 47.4 50.6 50.0 SOURCE: Automotive News Almanac, 1978, 1979; Consumer Reports (annu~il auto issue 1978, 1980). transaction price by about $30. This is the basic effect. In the case of characteristics that are either present or absent (e.g., diesel engines) the basic effect is the effect of having the characteristic. To find out whether the basic effect is a large or small one, we compare it to the typical difference in transaction prices, which in 1979 was $832. Thus, the basic effect of VOLWT , ~ ~ mu wan equal To '.o percent or the typical difference in the transaction price in 1979. We call this the relative effect. It may be useful to restate our basic hypothesis in terms of the variables in the analysis. In the price analysis the issue of competitive significance is essentially a question of the size and magnitude of the basic and relative effects. If a given technology characteristic, e.g., front-wheel drive, is a positive factor in competition, we would expect models with that characteristic to be higher priced. Thus, the sign of the basic effect should be positive; of course, an important factor will have-a larger relative effect. Table 8.3 contains the basic effect and the relative effect of , ~ q ~ ~ . _ . ~ ~ , . ~ .. . . ..
129 technological characteristics for 1977 and 1979. The results are striking. A line-by-line comparison of the effects of the various characteristics reveals a sharply different pattern of market valuation in these two years. While MPG was a positive impact in 1977, its effect in 1979 is swamped by the technology variables, and the estimated fuel-economy effect turns negative. The negative effect of MPG is puzzling but may reflect collinearity with the technology variables. Collinearity may also affect estimates of parameters on the technology variables. This seems to be especially true for VOLWT (package efficiency), where a large standard error precludes any strong conclusions about the sign of the price effect in either 1977 or 1979. For FWD (front- wheel drive) and AGE, however, collinearity does not seem to be a serious problem. If, for example, we drop the variables for diesel engine, age, and package efficiency (VOLWT) from the analysis, leaving only front-wheel drive as a measure of technology, we obtain basic and relative effects of front-wheel drive very similar to those reported in Table 8.3. And even if we reduced the 1979 F WD coefficient by two standard deviations, the resulting estimate (1 1 1 ) is still positive. In general, we find clear differences between the results for 1977 and those for 1979. TABLE 8.3 Basic and Relative Effects on Price Discounts of Performance and Technology Characteristics, 1977 and 1 979a 1977 1979 Basic Basic Effect Relative Effect Relative (current Effect (current Effect Characteristics dollars) (percentage) dollars) (percentage) Performance Characteristics Fuelefficiency (MPG) 192 21.3 -67 -8.1 Driving range (RNG) 5 0.6 65 7.8 Package efficiency -57 -6.3 30 3.6 (VOLWT) Technology Characteristics Engine type (DIESEL) -853 -94.6 302 36.3 Drive train (FOOD) -316 -35.0 417 50.1 Model age (AGE) -40 -4.4 -85 10.2 Market Segment Subcompact (SUB) - 174 - 19.6 313 37.6 a Effects for 1977 and 1979 are based on the coefficients given in equation 6 of Table C.3 in this volume.
130 The evidence suggests, for example, that consumers were willing to pay a premium for a diesel engine over and above premiums for greater range and greater fuel efficiency. In contrast, consumers in 1977 heavily discounted cars with front- wheel drive and diesel engines. Moreover, even the performance characteristics valued in 1979--range, package efficiency--had small or negative effects in 1977. While the market placed a high value on fuel efficiency in 1977, the performance and technology characteristics associated with market success in 1979 had little market appeal in 1977. If features that were considered inno- vative in 1979 had been introduced in 1977, they would have been greeted with above-average discounts. Introduction in such a con- text could only be interpreted as an attempt to force the market, for it seems clear that innovation, at least as defined in 1979 terms, was not valued in 1977. Perhaps the clearest indication of the value of innovation in 1977 and 1979 is the effect of model age. In 1979 the basic effect of a typical difference in age (2.3 years) was to lower the price by $85. Thus, newer cars received a premium even after controlling f or other attributes. Since performance and technology are already accounted for, the result implies that newness per se carried a premium in 1979. The effect of age in 1977 has the same sign but is much smaller and, as Appendix C shows, is not statistically significant. The results on the age variable in 1979 stand in contrast to the patterns of competition that have prevailed in the auto industry during most of the postwar era. Older models tend to be debug- ged, refined, and developed to meet a relatively stable set of ~~ User aemanas. ~ major model change may introduce new features and above-average performance, but if innovation is not itself valued the effects of bugs and introduction problems are likely to outweigh any value attached to the new features. This situation reversed in the late 1970s. The notion that market valuation of performance and technol- ogy was different after the oil shock of 1979 can be examined more rigorously using standard statistical tests. The null hypoth- esis in this connection is that the effects in the two years are identical. Appendix C provides technical details, but the statisti- cal tests confirm the apparent differences in Table 8.3. We are able to reject with a high degree of confidence the hypothesis that the effects are equal. In short, the evidence suggests that market valuation changed sharply after the oil shock of 1979, with technology playing a much more critical role. ~ ~ ~ . ~ .
131 COMPARISON OF SALES PER MODEL DURING 1977 AND 1979 The effects of performance and of technology characteristics on model sales in 1977 and 1979 are examined in Table 8.4. In con- trast to the results for price discounts, we find remarkable stability in the pattern of effects for 1977 and 1979. For most variables the direction of the effects are identical, and in several cases the orders of magnitude are roughly comparable. Even where there are differences the evidence in Appendix C suggests that little should be made of these results. The quality of the statistical evidence is poor, and the estimated effects are not statistically different from one another. Thus, while the more negative effects of diesel and front- wheel drive cast doubt on the hypothesis, and the evidence on age supports it, these changes are in fact more apparent than real. The data available provide no evidence that technology (as defined TABLE 8.4 Basic and Relative Effects on Sales of Performance and Technology Characteristics, 1977 and 1979a 1977 Characteristics Basic Effect (thousands of vehicles) 1979 Relative Effect (percentage) Basic Effect (thousands of vehicles) Relative Effect (percentage) Performance Characteristics Fuel efficiency (MPG) 35.0 37.5 10.1 10.6 Driving range (RNG) -46.2 -49.5 -6.6 -6.9 Package efficiency 32.9 35.2 36.5 38.4 (VOLWT) Technology Characteristics Engine type (DIESEL) -0.6 -0.6 -97.8 - 102.8 Drive train LEWD) -23.4 -25.1 -53.1 -55.8 Age (AGE) 4.9 5.2 -1.8 1.9 Repair Record Much worse than average (REP 1) Much better than average (REP 2) Market Segment Subcompact (SUB) 7.9 8.5 -61.3 -64.5 82.2 88.0 65.2 68.6 -84.8 94.0 3.0 3.2 a Effects are based on the coefficients given in equation 3 of Table C.4 in this volume.
132 here) had a positive effect on sales in either year or even that the sales relationships shifted dramatically. It may be that market shares are more stable than prices and that most of the effects of market shifts in 1979 were felt on the price side of the market. The results of the price analysis provide evidence that technology was an important aspect of competition in the 1979-1980 market. Compared with the situation in 1977, we find that technology characteristics and innovation were highly valued. At least as far as market premiums are concerned, changes in the market created an incentive for innovation consistent with the hypothesis advanced at the beginning of this chapter. In terms of their effect on prices, technology became more visible and competitively significant. N OTES 1. There is a large literature on the estimation of hedonic price equations. A review of the literature and an application can be found in Toder ( 1978~. 2. In effect what is required is a structural model of the implicit market for characteristics. With appropriate exogenous variables and exclusion restrictions, the model can be identified and estimated using one of several simultaneous equation methods. 3. It is obvious that this is a strong assumption, since the firm's pricing policy may attempt to estimate consumer valuation. Thus, the mode is only an approximation. 4. Although identification of estimated discounts effects depends on the assumption that list prices are based on standard costs and standard markups, the change in the estimated effects between two periods is likely to be more affected by demand considerations since relative costs of characteristics tend to be more stable. 5. This allows us to identify the effect of the technc~lc,av- holding performance constant. Of 6. In statistical terms, "typical difference" is the standard deviation of the variable in question. 7. Evaluation of the results should be tempered by the fact that the standard errors (in Appendix C of this volume) are sizeable in some cases. The lack of precision precludes strong conclusions for some variables. However, the overall pattern of effects is what is important, and it appears that overall the two patterns are different. 8. Discount equations with average repair records also were estimated, but the results were insignificant and were not reported in the test; see Appendix C of this volume.
