Selection of car attributes

For empirical applications of the hedonic price model, it is important to decide what product attributes the regression should entail to appropriately explain the

7Tables2.8and2.9give an overview of the selected models for gasoline and diesel cars, respectively, with the number of products and the average vehicle prices.

relationship between the price of a good and its characteristics. The model-building strategy in terms of the variable selection technique in this paper is based on the engineering background of the automotive industry, the quality of the available data, the car characteristics that are cited as important for buyers in industry overview reports⁸ and that have been used in previous studies, and various statistical criteria for a model fit (e.g., C_p, information criteria, and AdjustedR²).

The primary focus of this paper is the parameter estimate for fuel economy used in a subsequent analysis. The ADAC data provide three measures of fuel economy – city, highway, and weighted-average among city and highway values. In this paper, the latter measure is considered. From a technological perspective, however, fuel economy is strongly related to other car characteristics. This interdependence leads to a multicollinearity problem and, thus, to highly unstable parameter estimates and imprecisely estimated implicit prices. To overcome the strong interdependence between car attributes, many authors have proposed to include a variable that represents only one aspect of either fuel economy or vehicle performance (e.g., Agarwal and Ratchford, 1980). For example, Uri (1988) advises against any inclusion of the fuel economy variable, whereas Gramlich (2008) includes two different specifications of the fuel efficiency - miles per gallon (MPG) as a proxy for all other (“negative”) product qualities (“higher MPG is strongly associated with lower other quality”, p. 7) and the price of fuel divided by miles per gallon ($PM) as a measure of fuel economy itself. The present paper, however, undertakes another approach. Following the engineering literature, in which one may find a value of a power-specific fuel consumption (e.g., Van den Brink and Van Wee, 2001;Sprei et al., 2008), this paper considers a measure of fuel efficiency that is defined as a product of fuel economy with some indicator of a car’s performance.

In general, fuel efficiency refers to the amount of fuel necessary to produce a useful service output (Patterson,1996). A better value of fuel efficiency means that less fuel is needed for the same amount of output. Service output in the car example can be represented by various variables for car performance (e.g., horsepower, kW, power output per liter, etc.). In this paper, we follow previous studies and define fuel efficiency as a product of fuel economy and horsepower.⁹ This measure allows us to control for car performance while recovering the relationship between vehicle prices and fuel economy of a theoretically plausible direction. As can be seen in Table 2.1, the fuel economy of vehicles increases over model years but has a negative correlation with car prices, as shown in Table 2.2. We also see that fuel economy is highly correlated with various measures of car performance and engine characteristics. Advocated from a technology perspective, this pattern reflects

8The industry overview reports can be found, for example, athttp://www.dat.de.

9Other measures of car performance are highly correlated with horsepower and consequently yield statistically similar estimation results.

the fact that heavier and more powerful cars cost more but also consume more fuel. Adjusted by the car performance, however, the expected positive relationship between the vehicle price and fuel economy is restored.

Table 2.1: Fuel prices, car prices, and fuel efficiency over years

Diesel Gasoline

2011 2012 2013 2011 2012 2013

Fuel price Mean 1.39 1.42 1.34 1.50 1.54 1.47

SD 0.03 0.04 0.02 0.01 0.03 0.02

Car price Mean 29321.87 28485.04 28212.85 27352.10 27692.72 26572.58 SD 6936.29 6646.43 6253.51 7992.11 8287.36 7956.5

Fuel economy Mean 19.99 21.18 21.51 14.97 15.64 16.71

SD 2.51 2.75 3.26 1.93 2.01 2.32

Fuel efficiency Mean 2867.51 3017.62 3148.05 2369.69 2529.35 2682.21

SD 595.79 681.83 711.78 562.56 681.34 667.31

Forced induction Mean 1 1 1 0.71 0.75 0.86

SD 0 0 0 0.46 0.43 0.35

N 217 233 231 177 228 227

NOTE: Fuel prices are in 2010eper liter; car prices are in 2010e; fuel economy is in km/l. Fuel efficiency is defined as (fuel economy×horsepower). Values for forced induction are shares of the technology within all cars started being produced in a particular model year based on the ADAC data.

On average, diesel cars consume less fuel per unit distance than otherwise compara-ble gasoline vehicles. For example, a car from the compact class with 140 HP and manual transmission consumes, on average, 6.26 liter of fuel per 100 kilometers (≡16.17 km/l) in the case of a gasoline engine and only 4.93 l/100 km (≡ 20.51 km/l) with a diesel engine. However, the fuel efficiency of gasoline cars might be significantly improved by the use of forced induction in form of a turbocharger or a supercharger – a gasoline car with similar characteristics but with forced induction achieves 18.31 km/l, an improvement of 13%. The new technology also increases the price of a car. Without accounting for forced induction, gasoline cars are cheaper than diesel cars, but both types are priced similarly when they feature forced induction (Table 2.3). This phenomenon can be explained by the relative novelty of this technology applied to gasoline engines compared to diesel engines and by a relative gain in a car power under forced induction. Despite a relatively higher vehicle price, the share of gasoline cars with forced induction in the supply as a whole has been increasing over time. This finding leads to the conclusion that consumers might progressively value this technology.

Table 2.2: Correlation coefficients for a subset of vehicle attributes

Price FC FE FEff HP Displ Weight

Car price, 2010e 1 0.29 -0.26 0.67 0.79 0.76 0.70 Fuel consumption, l/100km 0.29 1 -0.97 -0.20 0.52 0.23 0.21 Fuel economy, km/l -0.26 -0.97 1 0.23 -0.48 -0.19 -0.20

Fuel efficiency 0.67 -0.20 0.23 1 0.72 0.67 0.34 CO2 emissions, g/km 0.39 0.96 -0.93 -0.12 0.56 0.37 0.38

Performance Characteristics

Horsepower (metric) 0.79 0.52 -0.48 0.72 1 0.74 0.44 Power, kW 0.79 0.52 -0.48 0.72 1 0.74 0.44 Acceleration, seconds -0.69 -0.32 0.31 -0.74 -0.87 -0.61 -0.24 Speed maximum, km/h 0.76 0.36 -0.34 0.76 0.91 0.65 0.35

Engine Characteristics

Displacement, cm³ 0.76 0.23 -0.19 0.67 0.74 1 0.55 Fuel Type (Gasoline = 1) -0.10 0.69 -0.70 -0.34 0.19 -0.26 -0.30 Forced induction (“yes”= 1) 0.30 -0.29 0.29 0.39 0.18 0.08 0.26 Transmission (Automatic = 1) 0.36 0.21 -0.22 0.12 0.25 0.24 0.20

Size Characteristics

Weight, kg 0.70 0.21 -0.20 0.34 0.44 0.55 1 Length, cm 0.49 0.19 -0.20 0.15 0.26 0.31 0.71

Width, cm 0.36 0.15 -0.14 0.13 0.21 0.19 0.63 Height, cm -0.28 0.07 -0.09 -0.41 -0.30 -0.21 0.13 NOTE: Reported are the Pearson correlation coefficients for continuous variables and the tetra-choric correlation coefficients for dichotomous variables. All values are statistically significant, with thep <0.01 unless otherwise stated; fuel efficiency is defined as (fuel economy×horsepower).